Browse Topic: Level 2 (Partial driving automation)
Autonomous vehicles utilise sensors, control systems and machine learning to independently navigate and operate through their surroundings, offering improved road safety, traffic management and enhanced mobility. This paper details the development, software architecture and simulation of control algorithms for key functionalities in a model that approaches Level 2 autonomy, utilising MATLAB Simulink and IPG CarMaker. The focus is on four critical areas: Autonomous Emergency Braking (AEB), Adaptive Cruise Control (ACC), Lane Detection (LD) and Traffic Object Detection. Also, the integration of low-level PID controllers for precise steering, braking and throttle actuation, ensures smooth and responsive vehicle behaviour. The hardware architecture is built around the Nvidia Jetson Nano and multiple Arduino Nano microcontrollers, each responsible for controlling specific actuators within the drive-by-wire system, which includes the steering, brake and throttle actuators. Communication between these components is facilitated through the CAN protocol, which ensures accurate and reliable data transfer essential for real-time decision-making. AEB achieves precise emergency braking and enhances driver comfort through the use of PID controllers, while ACC leverages radar data to maintain a safe distance from the vehicle ahead. LD employs the Hough Transform algorithm for accurate road edge detection. Furthermore, a trained neural network within the system identifies and responds to traffic signals, signage, pedestrians and vehicles. The camera interfaces directly with the Jetson Nano, while radar data is shared with the IMU through a dedicated CAN bus. This integrated approach represents a significant advancement in autonomous vehicle control, thus contributing to enhanced safety, comfort and reliability for both drivers and passengers. This software architecture is designed based on aBaja 2024 competition and according to its rules, regulations, requirements and specifications, the controllers and simulations were designed.
The automaker's recall of its Full Self Driving Beta leaves a significant dent in automated driving's credibility. On February 16, 2023, the National Highway Traffic Safety Administration announced that Tesla had voluntarily agreed to recall 362,758 Model S, Model X, Model 3 and Model Y vehicles - the entire parc of Tesla models fitted with the beta version of the company's Full Self-Driving (FSD) Beta software. The NHTSA cited FSD's failure to safely operate Tesla vehicles in a variety of common driving situations - while many industry sources contended the recall was proof Tesla no longer could stay one step ahead of the sheriff regarding its insinuations about FSD's capabilities. The tension about Tesla's automated-driving features had been building. In the summer of 2021, the NHTSA started investigating several crashes in which Teslas operating with the Autopilot ADAS system (standard on all models) struck parked emergency vehicles. A subsequent NHTSA report said there were 273 Autopilot-related crashes that year.
ADAS and HMI development are new applications for simulation solutions. The concept of designing, engineering and manufacturing a new vehicle without physical prototypes is typically viewed as either impractical or mythical. Even as virtual development processes have become increasingly capable, experts maintain that hard prototypes are still needed to validate the fidelity of virtual models. But “zero prototypes” is more than a slogan at one of the top providers of real-time simulation and driving simulator solutions. For VI-grade, zero prototypes are a crusade.
Mercedes-Benz developed an in-house computer operating system to join an all-new vehicle platform architecture to enhance automated driving, OTA updates and other features. Mercedes-Benz revealed in late February that it is developing its own computer operating system, dubbed MB.OS, which it said will be standardized across the company's entire model portfolio when deployment begins “mid-decade” in concert with the introduction of the equally new Mercedes Modular Architecture (MMA) vehicle platform. The MB.OS will have full access to all vehicle domains, including infotainment, automated driving, body and comfort, vehicle dynamics and battery charging. Based on a chip-to-cloud architecture, the company asserted MB.OS “is designed to connect the major aspects of the company's value chain, including development, production, omni-channel commerce and services - effectively making it an operating system for the entire Mercedes-Benz business.” The MB.OS architecture is completely updateable to enable rapid product upgrades and is “deliberately open for selected partners,” the company said. It also discreetly added that the new system is planned to “enhance customer lifetime value” by facilitating customer purchases such as content, services and vehicle functionalities. That strategy has been shown by competing automakers to enable certain vehicle features, for example, on a subscription-type basis.
One chip, multiple benefits. That's the claim made by U.S. semiconductor company Qualcomm Technologies Inc. about its new, scalable system-on-a-chip (SoC) product family, called Snapdragon Ride Flex. Unveiled at CES2023 and due to enter the market in early 2024, Snapdragon Flex is the auto industry's first scalable family of SoCs that can run a digital cockpit and ADAS features simultaneously, according to the company. Snapdragon Ride Flex is the latest member of the Snapdragon SoC family. Qualcomm's first-generation Ride Platforms are currently available in commercialized vehicles. Newer generations, which include the Ride Vision stack that can handle ADAS applications, are being tested by Tier 1s. They are expected to arrive on MY2025 vehicles from various OEMs, according to Qualcomm.
“The future happened yesterday” is an appropriate description of the rapid pace of development in automated-driving technology. The expression may be most accurate in sensor tech where, for most OEMs (except Tesla thus far), radar and lidar increasingly are considered an essential duo for enhanced automated driving beyond SAE Level 2, and of course for full Level 4-5 automation. Current lidar is transitioning from electro-mechanical systems to solid-state devices. Considered by industry experts to be the technology's inevitable future, Lidar 2.0 is next-generation 3D sensing that is software-defined, solid-state and scalable. Engineers believe those capabilities will make lidar ubiquitous by reducing costs, speeding innovation and improving user experience.
Every new industry sector goes through a consolidation process where the strongest survive, and so it is with automated and autonomous driving technologies. The recent shuttering of Argo AI, one of the autonomous-vehicle industry's leading tech companies, by Ford and Volkswagen might come as a surprise to commuters in San Francisco and in Phoenix, Arizona. Those who regularly use the robotaxi services of GM-backed Cruise Automation and Alphabet's Waymo see these and other AVs under development during their daily travels. On public roads. Every day. Indeed, Argo AI's demise (which insiders said was mainly due to friction among Ford and VW) and difficulties at other startups including AV pioneer Aurora, have highlighted the engineering challenges of safely achieving SAE Level 4 driving automation, while reinforcing AV critics. But as Guidehouse Insights' leading e-Mobility analyst Sam Abuelsamid notes in his Navigator column on page 3, the AV sector's leaders appear to be moving out ahead of the pack. Cruise has expanded its robotaxi service into Austin, Texas, with operations in more cities planned in 2023. The company's CEO Kyle Vogt claims Cruise will reach $1B in revenue by 2025. GM CEO Mary Barra is a believer.
As the level of automation is increasing, there is more sensing, processing of complex algorithms and actuation in the system. The Safety of intended functionality (SOTIF) becomes more and more relevant that address the functional insufficiencies or performance limitations of Autonomous functions. The functional insufficiencies/performance limitations can lead to undesired behaviors of the vehicle function for e.g., the system intervenes when there are no critical situations due to False positive scenarios which may lead to undesired braking, or the system does not react in a critical situation due to false negative scenarios which may lead to no braking when it is required to brake. To address these situations in the operational system, we develop SOTIF compliant system by identifying SOTIF risks and developing suitable measures to mitigate the identified risks. It is also necessary to Validate the system in right vehicle environment to confirm all the mitigation measures are effective and there is no unreasonable risk due to undesired behaviors. The validation target is generally specified in terms of the number of false positives that is acceptable during the endurance run for function being tested. In a Level 2 automated system, the focus is mainly to address and fix the false positives. Fixing the false positives is mostly at the cost of true positives (system intervenes in a critical situation). When the level of automation increases, the system becomes more responsible than the driver, then False negatives becomes increasingly relevant. In this paper, we describe how the false positives scenarios has been addressed as part of SOTIF, what are the challenges here, the challenges in addressing the False negative and the bigger and future challenge to design an ideal system with no/minimum false positive, no/minimum false negative and more true positives.
GM's latest SAE Level 2 system will enable hands-free operation in “95% of driving scenarios.” GM recently announced the next generation of its hands-free Super Cruise advanced driver-assist system (ADAS), upping the label to “Ultra Cruise.” GM claims the Ultra Cruise system, expected to appear first on Cadillac models in 2023, will ultimately enable hands-free driving on all paved public roads in the U.S. and Canada in “95% of driving scenarios.” At launch next year, the system is expected to cover more than 2 million miles of roads, with the capacity to grow to more than 3.4 million miles. Describing the system as a “door-to-door hands-free driving experience,” GM claims owners of Ultra Cruise-equipped vehicles will be able to travel hands-free across nearly every road, including highways, city and subdivision streets, along with paved rural routes. GM noted that its system has been developed completely in-house (via collaborating teams based in Israel, the U.S., Canada, and Ireland). Ultra Cruise will co-exist within the company's lineup with Super Cruise, which will become more widely available on mainstream vehicles, with Ultra Cruise reserved for more premium entries.
In this study we collect and analyze data on how hands-free automated lane centering systems affect the controllability of a hazardous event during an operational situation by a human operator. Through these data and their analysis, we seek to answer the following questions: Is Level 2 and Level 3 automated driving inherently uncontrollable as a result of a steering failure? Or, is there some level of operator control of hazardous situations occurring during Level 2 and Level 3 automated driving that can reasonably be expected, given that these systems still rely on a driver as the primary fall back. The controllability focus group experiments were carried out using an instrumented MY15 Jeep® Cherokee with a prototype Level 2 automated driving system that was modified to simulate a hands-free steering system on a closed track with speeds up to 110kph. The vehicle was also fitted with supplemental safety measures to ensure experimenter control. The lateral controllability study was carried out with a specific focus group of 61 participants representing a demographic unbiased population. Results suggest that driver hand position (hands on/off the steering wheel) significantly moderates a driver’s ability to control and recover from unexpected erroneous steering torque events, and that performance is impacted by torque profile (magnitude and ramp). Evidence suggests that drivers are able to control and recover from sudden unexpected steering torque events as high as 1.7 Nm and 5 Nm/sec ramp as long as a single hand is on the steering wheel in the vehicle under test. This paper discusses various lateral controllability metrics, and provides recommendations for application in hazard analyses of Level 2 or Level 3 automated driving systems.
As vehicles with SAE level 2 of autonomy become more widely deployed, they still rely on the human driver to monitor the driving task and take control during emergencies. It is therefore necessary to examine the Human Factors affecting a driver’s ability to recognize and execute a steering or pedal action in response to a dangerous situation when the autonomous system abruptly requests human intervention. This research used a driving simulator to introduce the concept of level 2 autonomy to a cohort of 60 drivers (male: 48%, female: 52%) of different age groups (teens 16 to 19: 32%, adults: 35 to 54: 37%, seniors 65+: 32%). Participants were surveyed for their perspectives on self-driving vehicles. They were then assessed on a driving simulator that mimicked SAE level 2 of autonomy. Participants’ interaction with the HMI was studied. A real-life scenario was programmed so that a request to intervene was issued when automation reached its boundaries while navigating a two-way curve road (TTC = 2.2 seconds). We found that at the time of the event, only 12% of participants kept their hands on the steering wheel. Only 64% of participants had their foot close to pedals. All participants who reacted within 0.65 seconds were able to avoid the crash. All participants who reacted after 0.9 seconds crashed. As a last construct, we looked at age and gender to understand how different participants behaved while vehicle automation was engaged.
The existing intelligent cruising assist system lacks comprehensive and objective test and evaluation scenarios. First, this paper analyzes the existing standard related to intelligent cruising assist system. Then, based on the current natural driving data, the paper analyzes the driving behavior of drivers, and extracts the information of speed, distance and road conditions in the Cruising scenario. Finally, this paper designed the test and evaluation scenarios of intelligent cruising assist system: the longitudinal control ability of one lane, the lateral control ability of one lane, the longitudinal and lateral control ability of one lane, Automatic lane change ability. The test and evaluation scenarios designed in this paper are used for the test and evaluation of intelligent cruising assist system at L2 level.
With the rapid development of artificial intelligence, autonomous driving technology will finally reshape an automotive industry. Although fully autonomous cars are not commercially available to common consumers at this stage, partially autonomous vehicles, which are defined as level 2 and level 3 autonomous vehicles by SAE J3016 standard, are widely tested by automakers and researchers. A typical Human-Machine-Interface (HMI) for a vehicle takes a form to support a human domination role. Although modern driving assistance systems allow vehicles to take over control at certain scenarios, the typical human-machine-interface has not changed dramatically for a long time. With deep learning neural network technologies penetrating into automotive applications, multi-modal communications between a driver and a vehicle can be enabled by a cost-effective solution. The multi-modal human-machine-interface will allow a driver to easily interact with autonomous vehicles, supporting smooth switching between human manual control and automation. However, unlike a steering wheel of vehicles, there is no normal or standard multi-modal human-machine-interface for autonomous vehicles. Moreover, unlike buttons and knobs, which cause little confusions in applications across different countries, multi-modal communications are affected by cultural nuances. Automotive Original Equipment Manufacturers (OEMs) can promote the typical human-machine-interface in different countries or automotive markets with little adaption, but OMEs need to adjust multi-modal human-machine-interface by taking into consideration cultural impacts, driving habits, social cognition and a traffic legal system. Design methodologies for human-machine-interface systems on different level autonomous vehicles are elaborated. The goal of multi-modal human-machine-interface in partially driving autonomous vehicles (SAE level 2 autonomous vehicles) and conditionally driving autonomous vehicles (SAE level 3 autonomous vehicles) is not only to mitigate a driver’s fatigue during driving, but also to keep certain amount of the driver’s engagement to ensure that the driver can take over control in a short time when the switching is necessary. The two sides of the design goal of multi-modal HMI systems are actually on a trade-off curve. Methodologies to optimize multi-modal communications to support HMI design are elaborated and compared in this paper.
In this work, we outline a process for traffic light detection in the context of autonomous vehicles and driver assistance technology features. For our approach, we leverage the automatic annotations from virtually generated data of road scenes. Using the automatically generated bounding boxes around the illuminated traffic lights themselves, we trained an 8-layer deep neural network, without pre-training, for classification of traffic light signals (green, amber, red). After training on virtual data, we tested the network on real world data collected from a forward facing camera on a vehicle. Our new region proposal technique uses color space conversion and contour extraction to identify candidate regions to feed to the deep neural network classifier. Depending on time of day, we convert our RGB images in order to more accurately extract the appropriate regions of interest and filter them based on color, shape and size. These candidate regions are fed to a deep neural network. In this paper, we focus on a region of interest (ROI) proposal method, which works to minimize false negative and false positive candidate regions that are then fed to the deep neural network for classification. This camera-only solution has applications for many levels of autonomy, from driver assistance technology (SAE Level 2) to fully automated vehicles (SAE Level 4).
Items per page:
50
1 – 28 of 28