Published Date: 2018-10-24
| Postmortem Analysis | |
|---|---|
| Timeline | 1. The software failure incident happened in 2018. |
| System | 1. Cruise cars' software struggled to identify whether objects on the road were moving or stationary, leading to hesitation and erratic braking [77324]. 2. The software failed to recognize pedestrians and mistakenly identified phantom bicycles, causing further issues with braking [77324]. 3. Cruise did not have a data-sharing collaboration with the San Francisco Fire Department, hindering the training of the cars to respond to fire truck sirens [77324]. 4. The open-source software robotics tools used by Cruise had delays that slowed messages from the car's sensors to the car's brain [77324]. |
| Responsible Organization | 1. Cruise and GM employees and executives [77324] 2. Autonomous vehicle technology experts familiar with Cruise [77324] |
| Impacted Organization | 1. General Motors Co and its Cruise self-driving car unit [77324] |
| Software Causes | 1. The software struggled to identify whether objects on the road were moving or stationary, leading to hesitation and erratic braking [77324]. 2. The software failed to recognize pedestrians and mistakenly identified phantom bicycles, causing braking issues [77324]. 3. The software did not have a data-sharing collaboration with the San Francisco Fire Department, hindering the training of the cars to respond to fire truck sirens [77324]. 4. Delays in the open-source software robotics tools used by Cruise slowed messages from the car's sensors to the car's brain [77324]. |
| Non-software Causes | 1. Lack of data-sharing collaboration with the San Francisco Fire Department to train the cars to respond to fire truck sirens [77324]. 2. Delays in messages from the car's sensors to the car's brain due to open-source software robotics tools used in development [77324]. |
| Impacts | 1. The software failure incident impacted General Motors Co and its Cruise self-driving car unit's ability to meet their 2019 goal of putting driverless cars on the road in a large scale way, as technical challenges, including difficulties in identifying moving objects, have caused delays and missed milestones [77324]. 2. The failure of the software to accurately identify whether objects on the road are moving or stationary led to Cruise cars hesitating and stopping while passing parked motorcycles or bicycles, as well as failing to recognize pedestrians and mistakenly detecting phantom bicycles, resulting in erratic braking behavior [77324]. 3. The software failure incident also highlighted issues with the open-source software robotics tools used by Cruise, which caused delays in transmitting messages from the car's sensors to its brain, impacting the overall performance of the self-driving cars [77324]. 4. The failure to establish a data-sharing collaboration with the San Francisco Fire Department hindered the training of the cars to respond to fire truck sirens, showcasing a limitation in the software's capabilities [77324]. 5. The software failure incident led to missed targets and milestones, such as recording the fewest number of times a human had to take control of a vehicle compared to rivals and achieving a million miles of testing per month, affecting the overall progress and performance of the autonomous vehicle technology [77324]. |
| Preventions | 1. Conducting more thorough testing and validation of the software to identify and address issues such as the difficulty in identifying moving objects on the road and recognizing pedestrians [77324]. 2. Implementing a more robust data-sharing collaboration with relevant entities like the San Francisco Fire Department to train the cars to respond to emergency situations [77324]. 3. Ensuring that the open-source software robotics tools used in developing the technology do not cause delays in processing sensor data, which can impact the car's performance [77324]. 4. Setting realistic milestones and targets for the development of the autonomous vehicle technology to avoid missing deadlines and facing setbacks [77324]. |
| Fixes | 1. Implement the next generation of hardware, software, and sensors to address issues with identifying moving objects and improve overall performance [77324]. 2. Enhance the software to accurately recognize pedestrians and differentiate between stationary and moving objects on the road to prevent erratic braking [77324]. 3. Establish a data-sharing collaboration with the San Francisco Fire Department to train the cars to respond to fire truck sirens [77324]. 4. Address delays in the open-source software robotics tools used for development to improve the speed of messages from the car's sensors to its brain [77324]. 5. Ensure that the Cruise system meets safety standards established by the automaker and shown to regulators before moving forward with the 2019 goal of offering rides to the public without safety drivers [77324]. | References | 1. Eight current and former GM and Cruise employees and executives 2. Nine autonomous vehicle technology experts familiar with Cruise 3. Michael Ronen, managing partner for SoftBank Investment Advisers and lead investor on the Cruise deal 4. San Francisco Fire Department spokesman 5. Four former Cruise employees who witnessed the problem with identifying moving objects 6. Nine other people familiar with Cruise's technology 7. GM President Dan Ammann 8. Honda executive Seiji Kuraishi 9. Industry source familiar with the matter 10. Paul Lienert, Joe White, and Ben Klayman from Reuters (additional reporting) [77324] |
| Category | Option | Rationale |
|---|---|---|
| Recurring | one_organization, multiple_organization | (a) The software failure incident having happened again at one_organization: - The article mentions that Cruise, the self-driving car unit of General Motors, has faced technical challenges, including difficulties in identifying moving objects on the road, leading to the vehicles hesitating and stopping unexpectedly [77324]. - Cruise's CEO, Kyle Vogt, acknowledged that there have been phases in development where systems didn't meet the requirements needed for launch, indicating ongoing software challenges [77324]. - Cruise and GM have missed certain milestones and targets, suggesting recurring software issues within the organization [77324]. (b) The software failure incident having happened again at multiple_organization: - The article highlights that Uber, a competitor of Cruise, had to overhaul its production timeline following a fatal crash involving one of its self-driving SUVs, indicating software challenges faced by multiple organizations in the autonomous vehicle industry [77324]. - Klaus Froehlich, a board member at BMW, expressed concerns about the industry wasting billions of dollars due to challenges in autonomous vehicle technology, implying that multiple organizations are facing similar difficulties [77324]. |
| Phase (Design/Operation) | design, operation | (a) The software failure incident related to the development phase of design can be seen in the challenges faced by General Motors Co and its Cruise self-driving car unit in identifying whether objects are in motion. The article mentions that Cruise cars struggle to identify whether objects on the road are moving or stationary, leading to hesitation and erratic braking behavior [77324]. (b) The software failure incident related to the development phase of operation can be observed in the article where it is mentioned that the driverless Cruise cars have failed to recognize pedestrians at times and have mistakenly identified phantom bicycles, causing the cars to brake erratically. Additionally, the software has not yet established a data-sharing collaboration with the San Francisco Fire Department, which is necessary for training the cars to respond to fire truck sirens [77324]. |
| Boundary (Internal/External) | within_system | (a) The software failure incident described in the articles is primarily within_system. The failure is attributed to technical challenges faced by General Motors Co and its Cruise self-driving car unit, such as the difficulty in identifying whether objects are in motion, struggles in recognizing pedestrians, and delays in the open-source software robotics tools used for development [77324]. These issues are internal to the system being developed and are hindering the progress of deploying driverless cars on a large scale by the set deadline. |
| Nature (Human/Non-human) | non-human_actions | (a) The software failure incident occurring due to non-human actions: - The article mentions technical challenges faced by GM and Cruise in developing their self-driving cars, including the difficulty of Cruise cars in identifying whether objects are in motion, resulting in hesitation and stopping while passing parked motorcycles or bicycles [77324]. - The software has also failed to recognize pedestrians and has mistakenly seen phantom bicycles, causing erratic braking of the cars [77324]. - Cruise's open-source software robotics tools used in developing the technology have delays that slow messages from the car's sensors to the car's brain [77324]. (b) The software failure incident occurring due to human actions: - The article does not specifically mention any software failure incident directly caused by human actions. |
| Dimension (Hardware/Software) | hardware, software | (a) The software failure incident occurring due to hardware: - The article mentions that the driverless Cruise cars struggle to identify whether objects on the road are moving or stationary, leading to hesitation and erratic braking. This issue is related to the hardware's sensors and their ability to accurately detect and interpret the environment [77324]. - It is highlighted that the open-source software robotics tools used by Cruise have delays that slow messages from the car's sensors to the car's brain, indicating a hardware-related bottleneck in the communication between sensors and processing units [77324]. (b) The software failure incident occurring due to software: - The article discusses how the software used by Cruise has failed to recognize pedestrians at times and has mistakenly identified phantom bicycles, causing erratic behavior in the cars. These issues point towards software-related challenges in object recognition and decision-making algorithms [77324]. - Additionally, the lack of data-sharing collaboration with the San Francisco Fire Department to train the cars to respond to fire truck sirens indicates a software-related limitation in integrating external data sources and adapting the software to respond appropriately to different scenarios [77324]. |
| Objective (Malicious/Non-malicious) | non-malicious | (a) The articles do not mention any malicious intent or actions contributing to the software failure incident. The challenges faced by General Motors Co and its Cruise self-driving car unit are related to unexpected technical difficulties, such as the struggle of Cruise cars to identify whether objects on the road are moving or stationary, failure to recognize pedestrians, and delays in the open-source software robotics tools used for development [77324]. These issues point towards non-malicious factors causing the software failure incident. |
| Intent (Poor/Accidental Decisions) | accidental_decisions | (a) The articles do not provide information about the software failure incident being related to poor decisions. (b) The software failure incident mentioned in the articles seems to be more related to accidental decisions or unintended mistakes rather than poor decisions [77324]. |
| Capability (Incompetence/Accidental) | development_incompetence | (a) The software failure incident occurring due to development incompetence: - The article highlights technical challenges faced by General Motors Co and its Cruise self-driving car unit, including the difficulty in identifying whether objects are in motion, leading to delays in putting driverless cars on the road at a large scale in 2019 [77324]. - Cruise cars struggle to identify whether objects on the road are moving or stationary, resulting in hesitation and erratic braking, as well as failing to recognize pedestrians and mistakenly detecting phantom bicycles [77324]. - The open-source software robotics tools used by Cruise have delays that slow messages from the car's sensors to the car's brain, indicating potential development inefficiencies [77324]. (b) The software failure incident occurring accidentally: - The article does not specifically mention any software failure incident occurring accidentally. |
| Duration | temporary | The software failure incident described in the articles seems to be more of a temporary failure rather than a permanent one. The articles mention that the Cruise self-driving cars are facing unexpected technical challenges, such as difficulty in identifying whether objects are in motion, struggles in recognizing pedestrians, and issues with phantom bicycles causing erratic braking [77324]. These challenges indicate that the failure is due to specific circumstances and technical limitations that can potentially be addressed through improvements in hardware, software, and sensors, as mentioned by Cruise CEO Kyle Vogt [77324]. Additionally, the article highlights ongoing efforts to enhance the technology and overcome these obstacles to meet safety standards and achieve the goal of launching the driverless cars on the road without drivers [77324]. |
| Behaviour | crash, omission, value, other | (a) crash: The article mentions instances where the software has failed to recognize pedestrians, seen phantom bicycles, and hesitated or stopped while passing parked motorcycles or bicycles, leading to erratic braking behavior [77324]. (b) omission: The software has struggled to identify whether objects on the road are moving or stationary, resulting in the vehicles hesitating and stopping while passing certain objects [77324]. (c) timing: There is no specific mention of timing-related failures in the articles provided. (d) value: The software has failed to correctly identify objects on the road, leading to instances where it has mistaken pedestrians and phantom bicycles, causing erratic braking behavior [77324]. (e) byzantine: The article does not explicitly mention the software behaving with inconsistent responses and interactions. (f) other: The software has faced technical challenges, including delays in processing messages from the car's sensors to the car's brain due to open-source software robotics tools used in development [77324]. |
| Layer | Option | Rationale |
|---|---|---|
| Perception | sensor, processing_unit, embedded_software | The articles provide information related to failures in the perception layer of the cyber physical system, specifically in the sensor, processing unit, and embedded software components: 1. **Sensor (a)**: - The article mentions that the driverless Cruise cars struggled to identify whether objects on the road were moving or stationary, leading to hesitation and erratic braking behavior [77324]. - It also states that the software at times failed to recognize pedestrians and mistakenly identified phantom bicycles, causing the cars to brake erratically [77324]. 2. **Processing Unit (c)**: - The open-source software robotics tools used by Cruise to develop the technology had delays that slowed messages from the car's sensors to the car's brain, affecting the overall performance [77324]. 3. **Embedded Software (e)**: - Cruise's CEO mentioned that the next generation of hardware, software, and sensors in the pipeline could help address issues related to object recognition and improve performance [77324]. - The article highlights that during the development phase, there were instances where systems did not meet the requirements needed for launch, indicating challenges in the embedded software development process [77324]. |
| Communication | unknown | The articles do not provide specific information about a software failure incident related to the communication layer of the cyber physical system that failed at either the link level or connectivity level. |
| Application | FALSE | Based on the information provided in the articles, there is no specific mention of the failure being related to the application layer of the cyber physical system due to bugs, operating system errors, unhandled exceptions, or incorrect usage. Therefore, it is unknown if the failure was specifically related to the application layer as described. |
| Category | Option | Rationale |
|---|---|---|
| Consequence | delay, non-human, theoretical_consequence | (a) death: There is no mention of any deaths resulting from the software failure incident in the provided article [77324]. (b) harm: The article does not mention any physical harm caused to individuals due to the software failure incident [77324]. (c) basic: The software failure incident did not impact people's access to food or shelter as per the article [77324]. (d) property: The software failure incident did not result in any direct impact on people's material goods, money, or data [77324]. (e) delay: The software failure incident did cause delays in the development and deployment of GM's driverless cars, impacting the company's timeline and milestones [77324]. (f) non-human: The software failure incident affected the performance of the driverless Cruise cars, leading to issues such as hesitation, stopping, and erratic braking due to the software's struggle to identify moving objects correctly [77324]. (g) no_consequence: The article does not mention any real observed consequences of the software failure incident [77324]. (h) theoretical_consequence: There were discussions about potential consequences of the software failure incident, such as the need to achieve safety standards before launching the driverless cars and concerns about wasting billions of dollars in the industry [77324]. (i) other: The article does not mention any other specific consequences of the software failure incident beyond those discussed in the options (a) to (h) [77324]. |
| Domain | transportation | The software failure incident discussed in the news article is related to the transportation industry. General Motors Co and its Cruise self-driving car unit were working on developing a robot taxi service to navigate the city streets of San Francisco. However, the project faced technical challenges, including issues with Cruise cars identifying moving objects, recognizing pedestrians, and responding to emergency vehicles like fire trucks. These challenges have led to delays in the deployment of driverless cars for commercial use [Article 77324]. |
Article ID: 77324