This paper proposes the utilization of a multi-mode ACC based on interpretable Finite State Machine (FSM) to address challenges in infrastructure and surrounding condition changes while meeting macroscopic safety and comfort requirements in traffic flow. Specifically, the paper designs and simulates a merge yield control mode, a dynamic speed change mode, and a safety and monitoring mode switching under defined state transitions with the traditional car-following mode. Advanced ACC algorithms have been applied to improve traffic efficiency and have demonstrated energy savings. Yet they have typically been deployed in a single-use case: car-following mode. In this mode, where the Autonomous Vehicle (AV) maintains an appropriate distance or time gap with the preceding vehicle, decelerating when the gap is small and accelerating when it is large, the system may struggle to guarantee safety and comfort in complex and variable driving scenarios. Although there exist mode-switching ACC and merge mode controllers, in which even involving latitude direction control have been proposed, their dynamics when driving alongside other controlled or human-driving vehicles under a Connected and Autonomous Vehicles (CAVs) traffic environment remains unclear. The paper includes results from an implementation that was successfully tested on the open road, and simulation results that show dampened disturbances from the mode-switching approach, compared to single-mode use of an ACC controller in the same scenarios.

This work introduces a multi-camera tracking dataset consisting of 234 hours of video data recorded concurrently from 234 overlapping HD cameras covering a 4.2 mile stretch of 8-10 lane interstate highway near Nashville, TN. Video is recorded in cooperation with Tennessee State Department of Transportation and its policies. The video is recorded during a period of high traffic density with 500+ objects typically visible within the scene and typical object longevities of 3-15 minutes. GPS trajectories from 270 vehicle passes through the scene are manually corrected in the video data to provide a set of ground-truth trajectories for recall-oriented tracking metrics, and object detections are provided for each camera in the scene (159 million total before cross-camera fusion). Initial benchmarking of tracking-by-detection algorithms is performed against the GPS trajectories, and a best HOTA of only 9.5% is obtained (best recall 75.9% at IOU 0.1, 47.9 average IDs per ground truth object), indicating the benchmarked trackers do not perform sufficiently well at the long temporal and spatial durations required for traffic scene understanding. Video data, scene information, and vehicle trajectories are made publicly available at i24motion.org.

Designing and validating controllers for connected and automated vehicles to enhance traffic flow presents significant challenges, from the complexity of replicating real-world stop-and-go traffic dynamics in simulation, to the intricacies involved in transitioning from simulation to actual deployment. In this work, we present a full pipeline from data collection to controller deployment. Specifically, we collect 772 km of driving data from the I-24 in Tennessee, and use it to build a one-lane simulator, placing simulated vehicles behind real-world trajectories. Using policy-gradient methods with an asymmetric critic, we improve fuel efficiency by over 10% when simulating congested scenarios. Our comprehensive approach includes reinforcement learning for controller training, software verification, hardware validation and setup, and navigating various sim-to-real challenges. Furthermore, we analyze the controller's behavior and wave-smoothing properties, and deploy it on four Toyota Rav4’s in a real-world validation experiment on the I-24. Finally, we release the driving dataset (Nice et al., 2021), the simulator and the trained controller (Lichtlé et al., 2022), to enable future benchmarking and controller design.

In this article, we assess the string stability of seven 2018 model yearadaptive cruise control(ACC) equipped vehicles that are widely available in the US market. Seven distinct vehicle models from two different vehicle makes are analyzed using data collected from more than 1,200 miles of driving in car-following experiments with ACC engaged by the follower vehicle. The resulting dataset is used to identify the parameters of a linear second order delay differential equation model that approximates the behavior of the black box ACC systems. The string stability of the data-fitted model associated with each vehicle is assessed, and the main finding is that all seven vehicle models have string unstable ACC systems. For one commonly available vehicle model that offers ACC as a standard feature on all trim levels, we validate the string stability finding with a multi-vehicle homogeneousplatoon experiment in which all vehicles are the same year, make, and model. In this test, an initial disturbance of 6 mph is amplified to a 25 mph disturbance, at which point the last vehicle in the platoon is observed to disengage the ACC. The data collected in the driving experiments is made available, representing the largest publicly available comparative driving dataset on ACC equipped vehicles.

Complex design considerations of Cyber-Physical Systems (CPSs) are greatly aided through the use of high-level abstractions in a Domain-Specific Modeling Language (DSML). A DSML environment ensures that models will be correct by construction, however syntax correctness does not imply dynamic safety. Presented is an approach to define a meta-metamodel for DSML design with built in closed loop verification feedback for model correction based on dynamic constraint violations. Usage of this framework during DSML design can benefit CPSs such as autonomous vehicles by ensuring that designs are safe, through codified expert knowledge and verification tool integration. A case study is shown using a simple hybrid controller modeling language and reachability verification tools.

Managing complexity while ensuring safely designed behaviors is important for cyber-physical systems as they are continually intro- duced to consumers, such as autonomous vehicles. Safety consid- erations are important as programming interfaces become open to experts and non experts of varying degrees. Autonomous vehicles are an example system where many domain experts must collab- orate together to ensure safe operation. Through the use of higher level abstraction, domain experts may provide verification tools to check dynamic behavioral constraints. Similarly, higher level mod- eling tools may generate lower level artifacts for a working system. With modeling and verification tools, smaller teams and potentially non-experts may program custom behaviors while ensuring a cor- rectly behaved system. A high level domain specific modeling language was created with a focus on non-domain experts. The domain consists of driving a vehicle through a set of known waypoints by connecting together multiple primitive motions in a sequence. Though constrained to simple motions, it is still possible to create a sequence to drive the vehicle unsafely. Model verification was implemented to check that the expected start and stop waypoints were correctly reached without driving the vehicle into unsafe regions. The language was then provided to a set of 4th year elementary school students to create unique paths. The models created by the students were then used to generate controller artifacts to operate a real autonomous vehicle on a soccer field.

Autonomous vehicles and other robotics systems are frequently implemented using a general-purpose programming language such as C++, and prototyped using domain-specific tool such as MATLAB/Simulink, and LabVIEW. Such an approach is not efficient when programming primitive motions of autonomous vehicles when considering important safety constraints, and when promoting the broad access to robotic systems through involvement of students and aspiring students who do not know conventional low-level programming languages. Aside from general-purpose programming languages, there are languages that are specifically designed to model autonomous vehicles, such as SHIFT, but these languages are typically for simulation purposes only. This experience report discusses the creation of a domain-specific language that allows for faster programming of autonomous vehicles while ensuring valid constraints will be met. This language generates code for multiple controllers that will operate alternatively to allow for fast and effective programming of vehicle trajectories using primitive motions. In addition to improving coding efficiency and reducing the number of programming errors, the language adds a level of abstraction so that autonomous vehicle behaviors may be generated by people with little knowledge of low-level details of the car's operation. Furthermore, this language ensures safe operation of the vehicle by enforcing a set of user-definable constraints on the output path. A main set of constraints that are applied to every generated path have been specifically chosen to enforce safe switching between controllers and prevent the planning of unsafe actions. A novel application of the language is its ability to permit users to add specific constraints for a particular path; these constraints are checked for validity after the main constraint check is performed.

Robotic systems have benefitted from standardized middleware that can componentize the development of new capabilities for a robot. The popularity of these robotic middleware systems has resulted in sizable libraries of components that are now available to roboti- cists. However, many robotic systems (such as autonomous vehicles) must adhere to externally defined standards that do not contain a large repository of components. Due to the real-time and safety concerns that accompany the domain of unmanned systems, it is not trivial to interface these middleware systems. However, previous attempts to do so have succeeded at the cost of ad hoc design and implementation. This paper describes a domain-specific approach to the synthesis of a bridge between the popular Robotic Operating System (ROS) and the Joint Architecture for Unmanned Systems (JAUS). The domain-specific nature of the approach permits the bridge to be limited in scope by the application's specific messages (and their attribute mappings between JAUS/ROS), resulting in smaller code size and overhead than would be incurred by a generic solution. Our approach is validated by tests performed on an unmanned vehicle with and without the JAUS/ROS bridge.

High speed robot locomotion is one of the most challenging problems in mobile robots. Fast robots push the limits of mechanical design, control and perception. The cheetah is an existence proof of what the authors term an Ultra-High Speed Locomotor, which can attain speeds of greater than 50 leg lengths per second and can cover 10 meters in a single gait cycle. In this paper, the energetics and dynamics of high-speed quadruped movement are analyzed. A leg design is presented that combines a novel hybrid actuator concept plus biarticular muscles to create a lightweight (17 Newton) leg capable of generating 90 Newtons of force. Static and dynamic test results show the promise of the hybrid actuator system in ultra high-speed locomotors.