Adaptive cruise control is currently one of the most widely accepted vehicle driver assistance technologies. However, the main challenge is still due to the uncertainty about other traffic participants' intended driving maneuvers such as unexpected lane changes. In this paper, graphical modeling and unsupervised learning techniques are utilized to obtain a model for inferring the intention of drivers in surrounding vehicles. In particular, real-world driving data collected on public roads with only standard production sensors is utilized to synthesize a hidden Markov model with the aid of unscented Kalman filtering and the expectation maximization technique. Exact probabilistic inference is used to obtain a posterior probability distribution over the driver's intention. The proposed inference framework can be utilized to faster detect lane changes when compared to commercial rule-based approaches. This, in turn, gives the adaptive cruise control system more reaction time which yields a potential for improving safety and ride comfort.

Deep learning-based approaches have been widely used for training controllers for autonomous vehicles due to their powerful ability to approximate nonlinear functions or policies. However, the training process usually requires large labeled data sets and takes a lot of time. In this paper, we analyze the influences of features on the performance of controllers trained using the convolutional neural networks (CNNs), which gives a guideline of feature selection to reduce computation cost. We collect a large set of data using The Open Racing Car Simulator (TORCS) and classify the image features into three categories (sky-related, roadside-related, and road-related features). We then design two experimental frameworks to investigate the importance of each single feature for training a CNN controller. The first framework uses the training data with all three features included to train a controller, which is then tested with data that has one feature removed to evaluate the feature's effects. The second framework is trained with the data that has one feature excluded, while all three features are included in the test data. Different driving scenarios are selected to test and analyze the trained controllers using the two experimental frameworks. The experiment results show that (1) the road-related features are indispensable for training the controller, (2) the roadside-related features are useful to improve the generalizability of the controller to scenarios with complicated roadside information, and (3) the sky-related features have limited contribution to train an end-to-end autonomous vehicle controller.

We propose in this paper an autonomous motion planning framework for companion robots to accompany humans in a socially desirable manner, which takes safety and comfort requirements into account. The overall framework consists of two parts: first, a novel parallel interacting multiple model-unscented Kalman filter (PIMM-UKF) approach is developed to simultaneously estimate human motion states and model mismatch, and then systematically predict the position and velocity of the human for a finite horizon. Second, based on the predicted human states, a nonlinear model predictive control (MPC) technique is utilized for the robot motion planning. The simulation results have demonstrated the superior performance in prediction using the PIMM-UKF approach. The effectiveness of the MPC planner is also shown by successfully facilitating the socially desirable companion behavior.

This paper proposes a novel algorithm to identify three inertial parameters: sprung mass, yaw moment of inertia, and longitudinal position of the center of gravity. A four-wheel nonlinear vehicle model with roll dynamics and a correlation between the inertial parameters is used for a dual unscented Kalman filter to simultaneously identify the inertial parameters and the vehicle state. A local observability analysis on the nonlinear vehicle model is used to activate and deactivate different modes of the proposed algorithm. Extensive CarSim simulations and experimental tests show the performance and robustness of the proposed approach on a flat road with a constant tire–road friction coefficient.

We consider a single-pursuer-multiple-evader pursuit-evasion game in which a team of evaders aims to delay the capture by a faster pursuer. We extend our previous open-loop formulation (and its solution) of the game to incorporate more realistic settings: a pursuer with uncertain position and evaders with limited turning rates. The formulation provides a guaranteed lower bound on the team survival time. The survival time performance of the proposed approach is evaluated through extensive simulations and compared to that of the existing approaches. It is shown to be highly effective even when the evaders can not detect the pursuer. A noticeable trend of potentially practical importance is that larger teams benefit more from an increase in turning rates than smaller teams.

Domestic grazing animals follow simple, scalable rules to assign themselves trajectories to cover a pasture. We explain how to adapt these rules for an information gathering system based on a realistic robot motion model and Kalman-filter based evidence grid that accounts for both bandwidth and sensor limitations. Our results show that this algorithm can meet or exceed the performance of state of the art field robotics systems, particularly when scalability and robustness to failure are required.

A robust method to estimate tire forces for a passenger vehicle with the Unscented Kalman Filter (UKF) is provided. Only standard vehicle sensors were used and no a priori knowledge of tire and road properties was required. The estimator uses the bicycle model and a random walk tire force model. The tire force estimates were compared to a CarSim reference model for combined slip maneuvers. The results showed a good overall tracking performance of the estimator. In addition, the UKF-estimator demonstrated a high convergence rate and good stability properties. The performed robustness studies showed that the estimator performs well even in the presence of disturbances such as changes in tire-road friction. This method enables a cost-effective and robust implementation for future real time vehicle applications.

A robust control design is proposed for the lane-keeping and obstacle avoidance of semiautonomous ground vehicles. A robust Model Predictive Controller (MPC) is used in order to enforce safety constraints with minimal control intervention. An uncertain driver model is used to obtain sets of predicted vehicle trajectories in closed-loop with the predicted driver's behavior. The robust MPC computes the smallest corrective steering action needed to keep the driver safe for all predicted trajectories in the set. Simulations of a driver approaching multiple obstacles, with uncertainty obtained from measured data, show the effect of the proposed framework.

Lateral tire deflection enables the estimation of the tire-road friction coefficient. Vehicle steering, such as driving on a curved highway, can influence the friction coefficient estimation. This paper demonstrates an algorithm to estimate the tire-road friction coefficient when a vehicle is steering. The relationship between the friction coefficient and the vehicle steering is derived through a tire brush model. The change of lateral velocity inside the tire-road contact patch is used in the algorithm along with the lateral deflection. The models for the lateral deflection and the change of lateral velocity are derived with the tire brush model, a simple tire model, and a parabolic lateral deflection model. Approximated tire slip angles are fed to the estimation algorithm to capture the change of the steering angle. This algorithm is evaluated in experiments with the steering of a test vehicle.

This paper presents the design of a novel active safety system preventing unintended roadway departures. The proposed framework unifies threat assessment, stability, and control of passenger vehicles into a single combined optimization problem. A nonlinear model predictive control (MPC) problem is formulated, where nonlinear vehicle dynamics, in closed-loop with a driver model, is used to optimize the steering and braking actions needed to keep the driver safe. A model of the driver's nominal behavior is estimated based on his observed behavior. The driver commands the vehicle, whereas the safety system corrects the driver's steering and braking actions in case there is a risk that the vehicle will unintentionally depart from the road. The resulting predictive controller is always active, and mode switching is not necessary. We show simulation results detailing the behavior of the proposed controller and experimental results obtained by implementing the proposed framework on embedded hardware in a passenger vehicle. The results demonstrate the capability of the proposed controller to detect and avoid roadway departures while avoiding unnecessary interventions.

The Collaborative Sensing Language (CSL) is a high-level feedback control language for mobile sensor networks (MSN). It specifies MSN controllers to accomplish network objectives with a dynamically changing ad-hoc resource pool. Furthermore, CSL is designed to allow the updating of controllers during execution (patching). This enables hierarchical control with simpler controllers at lower levels. The CSL Execution Engine contains the intelligence to allocate resources to tasks dynamically and adjust in real time to resource motion, this enables CSL controllers to be simple, intuitive and scalable. Experimental results show that the CSL Execution Engine performs these services with the addition of very little overhead.

A vision-based obstacle detection system for small unmanned aerial vehicles (UAVs) is presented. Obstacles are detected by segmenting the image into sky and non-sky regions and treating the non-sky regions as obstacles. The feasibility of this approach is demonstrated by using the vision output to steer a small unmanned aircraft to fly towards an obstacle. The experiment was first verified in a hardware in the loop (HIL) simulation and then successfully implemented on a small modified remote control plane using a large inflatable balloon as the obstacle.