Patent Application
20250045124
Traditionally, even as vehicles came to include more electronic components, each component operated independently of the others. For example, a vehicle may include power windows, power seats, automatic wipers, a radio, or some other type of electronic components providing vehicle functions-including combinations thereof. Each component may include its own processing circuitry to control the component and, depending on the type of component, may include wiring routed to different areas of the vehicle where different components are located. For example, power windows may require wires running from the switches and to the various windows being controlled. All the different components make it difficult, if not impossible, to update or expand features of the vehicle and wiring between components increases manufacturing complexity and costs for the vehicle.
The technology disclosed herein enables software-defined functions in a vehicle using asymmetric distributed compute nodes. In a particular example, a system includes one or more higher-performance compute nodes of the asymmetric distributed compute nodes and one or more lower-performance compute nodes of the asymmetric distributed compute nodes. The system further includes a communication backbone over which the higher-performance compute nodes and the lower-performance compute nodes communicate. The one or more higher-performance compute nodes execute first processes for performing a first portion of the software-defined vehicle functions for the vehicle and the one or more lower-performance compute nodes execute second processes for performing a second portion of the software-defined vehicle functions for the vehicle.
In another example, a method includes identifying first processes for execution on one or more higher-performance compute nodes of the asymmetric distributed compute nodes and identifying second processes for execution on one or more lower-performance compute nodes of the asymmetric distributed compute nodes. The method further includes directing the one or more higher-performance compute nodes to execute the first processes to perform a first portion of the software-defined vehicle functions for the vehicle and directing the one or more lower-performance compute nodes to execute the second processes to perform a second portion of the software-defined vehicle functions for the vehicle.
In a further example, an apparatus includes one or more microprocessors that execute first processes for performing a first portion of the software-defined vehicle functions for the vehicle. The apparatus also includes input and output interfaces connected to vehicle components of the vehicle and a network interface connected to a communication backbone to communicate with other compute nodes of the asymmetric distributed compute nodes. The other compute nodes execute second processes for performing a second portion of the software-defined vehicle functions for the vehicle.
In other examples, an apparatus performs the above-recited methods and computer readable storage media directs a processing system to perform the above-recited methods.
Vehicle functions may be handled by a single compute system, sometimes referred to as an Electronic Compute Unit (ECU). An ECU may use a single microprocessor, microcontroller, and/or a system-on-a-chip (SoC) to run a software operating system and applications to provide vehicle functions. The single computer interacts with other systems in the vehicle, including other single computers for vehicle components through external interfaces such as Universal Serial Bus (USB), universal asynchronous receiver-transmitter (UART), Controller Area Network (CAN) bus, Controller Area Network Flexible Data-Rate (CANFD) bus, Local Interconnect Network (LIN) bus, Ethernet, or some other communication convention-including combinations thereof. Typically, the ECU performs a different function than other computing components in the vehicle are capable of performing. When the ECU reaches its capacity for executing applications for the vehicle, no additional features defined by those applications can be added to the vehicle unless other features are removed or the software for the applications is otherwise pared down.
Using a network of distributed compute nodes, rather than a single ECU, increases the capacity in a vehicle for executing software that provides vehicle functions. While the ECU described above may be able to execute a software process, the other computing components are not designed to execute the process (e.g., are based on a different processing architecture) or there is otherwise no mechanism for executing the process on one of the other components (e.g., no connection or protocol needed to provide instructions to execute the process). Using a network of similar compute nodes enables the process to run on any one of the nodes. Nodes can, therefore, share the processing load and, in some situations, may serve as a backup node should another node fail. Moreover, not all nodes need have the same level of performance as many vehicle functions require relatively low processing power. The nodes may be asymmetric in their performance capabilities with some nodes having lower performance capabilities than other nodes. Processes that can be adequately executed on a lower-performance node may be offloaded to such a node thereby enabling a higher-performance node to execute other processes. Computing devices having lower performance typically cost less than their higher-performance counterparts, which leads to savings over installing all higher-performance nodes in a vehicle.
In operation, asymmetric distributed compute nodes 101-104 are asymmetric in the sense that the performance capabilities of each node are not the same. Higher-performance node 101 has higher performance capabilities than lower-performance nodes 102-104. Higher-performance node 101 may be higher performance than lower-performance nodes 102-104 due to having more processing cores, one or more faster processing cores, more memory (e.g., random access memory (RAM), cache memory, etc.), faster memory, faster communication circuitry, additional/different peripherals (e.g., a wireless modem), or some other computing component that enhances capabilities of higher-performance node 101 relative to capabilities of lower-performance nodes 102-104—including combinations thereof. While differing in performance, asymmetric distributed compute nodes 101-104 share a common computing architecture such that the instructions of processes 121-124 can execute on any of asymmetric distributed compute nodes 101-104. Vehicle 131 includes two tiers of nodes, but other examples may include one or more additional node tiers. For example, a medium-performance node may be present in a different implementation with the medium-performance node having higher performance capabilities than lower-performance nodes 102-104 but lower performance capabilities than higher-performance node 101. Also, in other examples, vehicle 131 may include different numbers of each tier of node. For instance, vehicle 131 may include more than one higher-performance compute node and/or some number other than three lower-performance compute nodes.
Asymmetric distributed compute nodes 101-104 communicate over communication backbone 111. Communication backbone 111 may include metallic, wireless, or optical links—or some combination thereof. Communication backbone 111 may use Internet Protocol (IP), Ethernet, optical networking, wireless protocols, communication signaling, or some other communication format—including combinations thereof. While communication backbone 111 is shown as having a direct link between each of asymmetric distributed compute nodes 101-104 to others of asymmetric distributed compute nodes 101-104, other network arrangements may be used. For instance, a network router or switch may be employed to direct traffic between asymmetric distributed compute nodes 101-104, which may enable each of asymmetric distributed compute nodes 101-104 to use one port to communicate with other nodes. In some of the alternative examples, network routing functionality may be included in one or more of asymmetric distributed compute nodes 101-104. For instance, lower-performance nodes 102-104 may each connect to higher-performance node 101 and higher-performance node 101 may handle the routing of traffic between lower-performance nodes 102-104. In some examples, communication backbone 111 may be used to connect other components of vehicle 131 outside of asymmetric distributed compute nodes 101-104.
Processes 121-124 execute to provide vehicle functions. The processes may be provided by a manufacturer of vehicle 131 or may be provided by a third party (e.g., the manufacturer of vehicle 131 may provide a development kit for third parties to develop processes for vehicle 131). As the processes are software instructions, the vehicle functions can be considered software defined. For example, while a separate controller may have been used previously to perform an automatic headlight function (e.g., a function that turns the headlights on or off depending on the light outside), that function can now be controlled via software rather than requiring a separate controller. Other example vehicle functions may include windshield wipers, exterior lighting, interior lighting, door locks/latches, steering wheel positioning, pedal positioning, mirror adjustment, radio/media handling (e.g., infotainment), engine or electric motor controls, battery charging, power seat adjustment, HVAC (Heating, Ventilation, and Air Conditioning), safety mechanisms (e.g., automatic braking, traction control, etc.), diagnostic tools, artificial intelligence (AI) algorithms, or any other type of function that a vehicle may include. In some examples, vehicle functions may be categorized into what the automotive industry commonly refers to as domains. Example domains may include powertrain, battery, electrical, infotainment, lighting, body, chassis, and advanced driver-assistance system (ADAS). While not shown, one or more of asymmetric distributed compute nodes 101-104 may be connected to vehicle components that are controlled by processes 121-124. For example, lower-performance node 103 may include an interface connected to brake lights of vehicle 131. Lower-performance node 103 can turn the brake lights off or on using the interface. For example, the interface may supply power to the brake lights when the brake lights should be turned on or may provide a signal to a power supply for the brake lights directing the power supply to power the brake lights when the brake lights should be turned on. Since vehicle components are located throughout vehicle 131, asymmetric distributed compute nodes 101-104 may similarly be distributed throughout vehicle 131 to connect with those components. Connecting a component to a nearby node lessens the amount of wiring needed had the component needed to connect to a compute unit located elsewhere in vehicle 131 (e.g., to a centralized ECU). The material savings increase as more components are connected to nearby nodes rather than a centralized compute unit and those savings translate into monetary cost savings. Likewise, reducing the complexity of wiring throughout the vehicle simplifies manufacturing and troubleshooting when a component does not work properly.
In some examples, other computing components may still exist in vehicle 131 beyond asymmetric distributed compute nodes 101-104 and may be connected to one or more of asymmetric distributed compute nodes 101-104 either via communication backbone 111 or a separate interface. For example, an engine or motor controller may remain separate from asymmetric distributed compute nodes 101-104 but may feed information to asymmetric distributed compute nodes 101-104 that may be used by one of processes 121-124 (e.g., a process that presents vehicle status information to a user).
After identifying the respective nodes of asymmetric distributed compute nodes 101-104 for processes 121-124, higher-performance node 101 is directed to execute processes 121 to perform a first portion of the software-defined vehicle functions for vehicle 131 (step 203). Likewise, lower-performance nodes 102-104 are directed to execute the processes 122-124 to perform a second portion of the software-defined vehicle functions for the vehicle (step 204). The control process mentioned above may direct the processes to be assigned to the identified nodes. Assigning a process to a node may include sending the program instructions for the process to the node for installation thereon. For example, higher-performance node 101 may receive the program instructions for one of processes 122 and may send those program instructions to lower-performance node 102 for execution. In other examples, the program instructions may be received at lower-performance node 102 from some other source.
Once processes 121-124 are distributed across asymmetric distributed compute nodes 101-104, the software-defined vehicle functions facilitated by processes 121-124 are provided like they would have been had processes 121-124 been executing on a single node, such as an ECU described above. Additionally, since more than one node is available to potentially execute a process, processes can be reassigned between asymmetric distributed compute nodes 101-104 (e.g., if the overall performance of processes 121-124 would be better with a different distribution of processes). Similarly, nodes may be backups to one another for certain processes. For example, processes 122 and processes 123 may include instances of the same process with one being active and the other on standby. Should something cause the active process to fail or otherwise not perform up to desired standards (e.g., a kernel crash on the node), then the standby process can be directed to takeover. A control process, such as the control process that handles the distribution of processes, may also handle the failure recognition and standby process activation responsive thereto. If a process does not have a standby process executing on another node, another node may still be used as a failover node by initiating the process on the other node after failure. Processes that are not critical to safety features of the vehicle may use this approach since non-safety processes can afford to be down while a new instance initiates. Lower-performance nodes 102-104 may also be able to backup higher-performance node 101 even if one or more of processes 121 would need to run in a reduced capacity to account for performance differences between the nodes.
In operation 300, a process is received for installation on a node of asymmetric distributed compute nodes 101-104 (step 301). The process may be received as an over-the-air update to processes executing on asymmetric distributed compute nodes 101-104, as is becoming more common, or may be received from some other source (e.g., a service technician uploading the process from a user device to one of asymmetric distributed compute nodes 101-104). In this example, the process does not identify a specific node on which it should be installed to asymmetric distributed compute nodes 101-104. As such, asymmetric distributed compute nodes 101-104 determine whether the received process needs to execute on higher-performance node 101 to perform properly (step 302). The received process may define node attributes, such as memory, processing resources (e.g., speed, number of cores, etc.), peripherals, or some other node attribute-including combinations thereof. If those attributes can only be supplied by a higher-performance compute node, then the process is assigned to higher-performance node 101. Specifically, asymmetric distributed compute nodes 101-104 install the process on higher-performance node 101 (step 305). Other manners of determining that a higher-performance compute node should execute the process may also be used. In some examples, more than one higher-performance compute node may be connected to communication backbone 111. In those examples, determinations similar to those made in step 303 described below may be performed to select one of the higher-performance compute nodes.
If the process can execute properly on a lower-performance compute node, then asymmetric distributed compute nodes 101-104 may use rules to select which of lower-performance nodes 102-104 the process should be assigned. In one example, a node of lower-performance nodes 102-104 is selected that is connected to a component controlled by the process (step 303A). For instance, if the process controls the automatic windshield wipers for vehicle 131, then a node may be selected to which the windshield wipers are connected. By selecting the node connected to the wipers, control of the wipers can be performed without having to send control messages over communication backbone 111 to another node. In another example, a load balancing algorithm may be performed to select one of lower-performance nodes 102-104 that balances the processing load among lower-performance nodes 102-104 (step 303B). Other manners of selecting nodes may also be used, such as random selection or round-robin selection. Additionally, in some examples, higher-performance node 101 may also be considered when determining which node to select. Even though a process may not require the additional performance of higher-performance node 101 to execute properly, the process can still execute on higher-performance node 101 and it may be preferable for the process to do so (e.g., based on load balancing determinations or higher-performance node 101 being connected to a controlled vehicle component).
Once selected, the process is assigned to the selected node by installing the process on the selected node (step 304). For example, the process may be installed on lower-performance node 102 to become part of processes 122. In some cases, load balancing may cause other processes to be moved to other nodes. For instance, the load balancing algorithm may determine that one or more processes on the selected node should be reassigned to maintain proper balance of processes 121-124 across asymmetric distributed compute nodes 101-104 when the received process is installed on the selected node.
Memory 402 may include random access memory, read only memory, magnetic disks, optical disks, and flash memory, as well as any combination or variation thereof, or any other type of storage media. In no examples would storage media included in memory 402, or any other computer-readable storage medium herein, be considered a transitory form of signal transmission (often referred to as “signals per se”), such as a propagating electrical or electromagnetic signal or carrier wave. Memory 402 may be configured to store processes for execution by microprocessors 401 to provide software-defined vehicle functions for vehicle 131.
Communication interface 404 includes components that connect to communication backbone 111, such as network cards, ports, radio frequency (RF), processing circuitry and software, or some other communication devices. Communication interface 404 may be configured to communicate over metallic, wireless, or optical links. Communication interface 404 may be configured to use Internet Protocol (IP), Ethernet, optical networking, wireless protocols, communication signaling, or some other communication format-including combinations thereof.
Component I/O 403 includes input and output interfaces for interacting with vehicle components external to asymmetric distributed compute nodes 101-104. Component I/O 403 may be different for each of asymmetric distributed compute nodes 101-104 depending on which vehicle components compute node 400 is configured to connect. Component I/O 403 may include analog, digital, pulse width modulation (PWM), or some other type of signal interfaces.
Additional peripherals 405 include other compute elements that compute node 400 may employ to provide software-defined vehicle functions. For example, additional peripherals 405 may include a cellular modem to connect asymmetric distributed compute nodes 101-104 to a cellular network over which processes can be received and data can be exchanged for providing software-defined vehicle functions. Other examples of additional peripherals 405 may include a graphics processing unit (GPU), audio amplifier, or some other type of peripheral.
Since compute node 400 represents an architecture for both higher-performance node 101 and lower-performance nodes 102-104, the exact component configuration may differ between node types. For example, compute node 400 for higher-performance node 101 may include more microprocessors 401 or higher performance processing cores within microprocessors 401, faster forms of memory 402, additional storage media in memory 402 (e.g., additional flash memory to store processes for execution), or other performance differentiators. Additional peripherals 405 may be different for different nodes as well (e.g., one node may include a cellular modem). In some examples, the additional performance of a higher-performance compute node may necessitate the inclusion of a more powerful power supply or more extensive cooling to handle the additional power used. The lack of such extra components in a lower-performance compute node may further enhance the cost benefits of lower-performance compute nodes relative to higher-performance compute nodes. Similarly, enabling processes to run on lower-performance compute nodes that use less power enhances energy efficiency of vehicle 131.
In operational scenario 500, user interface process 521 directs higher-performance node 101 to display an HVAC user interface at step 1 on display 501. The HVAC user interface may be part of a larger user interface for a user to interact with functions of vehicle 131. For example, the user interface may include sections (e.g., windows, tabs, screen zones, etc.) displaying information and user-selectable buttons for functions including HVAC, radio, media, navigation, gaming, vehicle settings, or any other type of function that a user may be able to control or receive information about in vehicle 131. The HVAC user interface of this example at least enables a user to select a temperature that for the HVAC system of vehicle 131 to hold automatically. The HVAC user interface may also include other options, such as manual fan control, air conditioning control, vent direction, or some other adjustable feature for a vehicle HVAC system-including combinations thereof. A user in vehicle 131 (e.g., the driver or a passenger) selects a desired temperature at step 2 for the interior of vehicle 131. For example, the user may touch up/down arrows of the HVAC user interface until a desired temperature is reached. The user input is relayed at step 3 to higher-performance node 101 via component I/O 403 of higher-performance node 101 to indicate the desired temperature.
Upon receiving the user input, user interface process 521 determines at step 4 that the input is intended for HVAC process 522. In this example, the HVAC user interface provided by user interface process 521 is associated with functions controlled by HVAC process 522 so user interface process 521 knows to pass the input to HVAC process 522. Accordingly, user interface process 521 transmits a message over communication backbone 111 at step 5 indicating the temperature setting to lower-performance node 104 on which HVAC process 522 is executing. The message may be in any format capable of being transmitted over communication backbone 111 and readable by HVAC process 522 at lower-performance node 104. HVAC process 522 may have been selected to execute on lower-performance node 104, as opposed to another of lower-performance nodes 102-104, because component I/O 403 of lower-performance node 104 is connected to HVAC components thermometer 502, fan 503, and heat pump 504. Although, in other examples, one or more of components 502-504 may be connected to another of asymmetric distributed compute nodes 101-104.
Thermometer 502 provides the current internal temperature within vehicle 131 at step 6 to HVAC process 522 via component I/O 403 of lower-performance node 104. HVAC process 522 determines how fan 503 and heat pump 504 should be controlled at step 7 to adjust the internal temperature of vehicle 131 to the set temperature. For example, if the current temperature is below the set temperature, then HVAC process 522 may determine to increase fan speed of fan 503 and trigger heat pump 504 to produce hot air. HVAC process 522 provides control signals to fan 503 and heat pump 504 at step 8 via component I/O 403 of lower-performance node 104 to achieve the set temperature adjustment. HVAC process 522 may continue controlling thermometer 502 and heat pump 504 once the set temperature is reached to ensure the current interior temperature remains at the set temp as indicated by thermometer 502.
While not shown in this example, HVAC process 522 may transmit a message indicating the current interior temperature over communication backbone 111 back to higher-performance node 101 so that user interface process 521 can present the current temperature on display 501.
In operational scenario 600, light sensor 601 provides an indication of low external light level at step 1 to light control process 621 via component I/O 403 of lower-performance node 104. The indication may explicitly inform light control process 621 that the light level has fallen below a threshold or may provide an amount of light that light control process 621 compares to a threshold amount of light. When the light level falls below a threshold amount of light, light control process 621 determines at step 2 that headlights 602 and taillights 603 should be turned on. In response to the determination, light control process 621 directs headlights 602 to turn on at step 3.
When headlights 602 are on, taillights 603 should also be on. Therefore, light control process 621 transmits a message over communication backbone 111 at step 4 with an instruction to turn on taillights 603. In some examples, light control process 621 may be aware that taillights 603 are connected to lower-performance node 102 and direct the message to lower-performance node 102. In other examples, light control process 621 may rely on another process of lower-performance node 104 to know that taillights 603 is connected to lower-performance node 102 and direct the message to lower-performance node 102 accordingly.
Upon receiving the message from lower-performance node 104, lower-performance node 102 controls taillights 603 to turn on at step 6 via component I/O 403 of lower-performance node 102. Thus, upon completion of step 6, both headlights 602 and taillights 603 have been turned on to account for the low-light situation external to vehicle 131. If the message from lower-performance node 104 is formatted and/or addressed in a manner that directs it to component I/O 403 of lower-performance node 102, then lower-performance node 102 may act as a zonal gateway between taillights 603 and lower-performance node 104. In other examples, a process of processes 122 executing on lower-performance node 102 may be configured to receive the message, determine that the message includes an instruction to taillights 603, and comply with the instruction by directing taillights 603 to turn on via component I/O 403 of lower-performance node 102. Similar routing circumstances may exist for information flowing from a component connected to component I/O 403 of lower-performance node 102 to other nodes. That is, sensor information may be automatically addressed/formatted for lower-performance node 102 to act as a zonal gateway directing the information to another node or a process executing on lower-performance node 102 may handle the routing of the information to another node.
In an example above, lower-performance node 102 may act as a zonal gateway. In some examples, communication backbone 111 may connect one or more zonal gateways to asymmetric distributed compute nodes 101-104. A zonal gateway includes component I/O 403 for vehicle components attached thereto and communication interface 404 to connect to communication backbone 111. Thus, a zonal gateway may provide an additional location in vehicle 131 to attach vehicle components while the processes that control those vehicle components execute on asymmetric distributed compute nodes 101-104.
In this example, distributed system controller 701 is configured to handle routing of instructions between nodes of asymmetric distributed compute nodes 101-104. Relying on distributed system controller 701 for routing, processes 121 do not need to be aware of which other nodes are connected to vehicle components controlled by thereby. When an instruction is received from processes 121 at step 1, distributed system controller 701 determines a destination node of asymmetric distributed compute nodes 101-104 at step 2 based on node mapping 721. Node mapping 721 is information that indicate which nodes include which vehicle components and which nodes are executing which processes. Distributed system controller 701 may update the mapping as needed when new processes are installed in asymmetric distributed compute nodes 101-104 or processes are moved between asymmetric distributed compute nodes 101-104. In this example, once distributed system controller 701 has determined a destination node for the instruction, distributed system controller 701 sends the instruction at step 3 to the destination node in a message over communication backbone 111 via communication interface 404 of higher-performance node 101.
Other nodes of asymmetric distributed compute nodes 101-104 may execute similar processes to distributed system controller 701 to route messages between nodes. Using the example from operational scenario 600, light control process 621 passes the instruction to turn on taillights 603 to a version of distributed system controller 701 executing on lower-performance node 104. That version of distributed system controller 701 determines based on a mapping like node mapping 721 that the instruction need be sent to lower-performance node 102 because lower-performance node 102 is connected to taillights 603. The message with the instruction is sent to lower-performance node 102 accordingly.
In some examples, distributed system controller 701 may be configured to route control instructions to vehicle components connected to the same node. For instance, upon receiving an instruction from processes 121, distributed system controller 701 may determine if the instruction applies to a vehicle component connected to higher-performance node 101 and route the instruction to component I/O 403 for that component. In other words, a process can be configured to generate control instructions for a vehicle component without having to consider to which node the vehicle component is connected. Distributed system controller 701 will ensure the instruction ends up at the right node. Additionally, distributed system controller 701 may be configured to handle incoming messages from communication backbone 111 and route the incoming message to the one of processes 121 or vehicle components to which they are directed (or similarly handle incoming information from component I/O 403 of higher-performance node 101). Distributed system controller 701, when executing on each of asymmetric distributed compute nodes 101-104, ensures the vehicle functions provided by processes 121-124 perform as seamlessly as they would if they were all executing on the same compute node. In some examples, the instances of distributed system controller 701 executing on lower-performance node 102 may include fewer features than distributed system controller 701 on higher-performance node 101 because distributed system controller 701 on higher-performance node 101 may handle all the logic for distributing processes 121 across asymmetric distributed compute nodes 101-104 (e.g., distributed system controller 701 on higher-performance node 101 may enable higher-performance node 101 to act as a master node of asymmetric distributed compute nodes 101-104). Other nodes may instead be the master in other examples and the master node may change should a previous master node fail.
A process like distributed system controller 701 may further handle determining the health status of other nodes. For example, each node may transmit heartbeat messages to distributed system controller 701 indicating that they are running. Should a node fail (e.g., as indicated by the lack of a heartbeat message for a period of time), distributed system controller 701 may perform operation 300 for the processes of the failed node to redistribute the processes across the remaining nodes.
Since asymmetric distributed compute nodes 801-806 include multiple higher-performance compute nodes, processes identified for execution on a higher-performance compute node will be distributed between higher-performance compute node 801 and higher-performance compute node 802. Logic similar to that used in step 303 of operation 300 above may be employed to determine which processes should execute on which of higher-performance compute nodes 801-802. Similarly, since asymmetric distributed compute nodes 801-806 include four lower-performance compute nodes, processes identified for execution on a lower-performance compute node will be distributed across four nodes rather than three, as was the case with implementation 100.
The included descriptions and figures depict specific implementations to teach those skilled in the art how to make and use the best mode. For teaching inventive principles, some conventional aspects have been simplified or omitted. Those skilled in the art will appreciate variations from these implementations that fall within the scope of the invention. Those skilled in the art will also appreciate that the features described above can be combined in various ways to form multiple implementations. As a result, the invention is not limited to the specific implementations described above, but only by the claims and their equivalents.CLAIMS
