Safety-aware Adversarial Inverse Reinforcement Learning (S-AIRL) for Highway Autonomous Driving

Inverse reinforcement learning (IRL) has been successfully applied in many robotics and autonomous driving studies without the need for hand-tuning a reward function. However, it suffers from safety issues. Compared to the reinforcement learning (RL) algorithms, IRL is even more vulnerable to unsafe situations as it can only infer the importance of safety based on expert demonstrations. In this paper, we propose a safety-aware adversarial inverse reinforcement learning algorithm (S-AIRL). First, the control barrier function (CBF) is used to guide the training of a safety critic, which leverages the knowledge of system dynamics in the sampling process without training an additional guiding policy. The trained safety critic is then integrated into the discriminator to help discern the generated data and expert demonstrations from the standpoint of safety. Finally, to further improve the safety awareness, a regulator is introduced in the loss function of the discriminator training to prevent the recovered reward function from assigning high rewards to the risky behaviors. We tested our S-AIRL in the highway autonomous driving scenario. Comparing to the original AIRL algorithm, with the same level of imitation learning (IL) performance, the proposed S-AIRL can reduce the collision rate by 32.6%.

Simultaneous Human-robot Matching and Routing for Multi-robot Tour Guiding under Time Uncertainty

This work presents a framework for multi-robot tour guidance in a partially known environment with uncertainty, such as a museum. In the proposed centralized multi-robot planner, a simultaneous matching and routing problem (SMRP) is formulated to match the humans with robot guides according to their selected points of interest and generate the routes and schedules for the robots according to uncertain spatial and time estimation. A large neighborhood search algorithm is developed to find sub-optimal low-cost solutions for the SMRP efficiently. The scalability and optimality of the multi-robot planner are first evaluated computationally under different environment sizes and numbers of humans and robots. Then, a photo-realistic multi-robot simulation platform was developed based on Habitat-AI to verify the tour guiding performance in an uncertain indoor environment. Results demonstrate that the proposed centralized tour planner is scalable, makes a smooth tradeoff in the plans under different environmental constraints, and can lead to robust performance with inaccurate uncertainty estimations (within a certain margin).

A Multi-Objective Optimization Approach for Multi-Vehicle Path Planning Problems considering Human-Robot Interactions

There has been unprecedented development in the field of unmanned ground vehicles (UGVs) over the past few years. UGVs have been used in many fields including civilian and military with applications such as military reconnaissance, transportation, and search and research missions. This is due to their increasing capabilities in terms of performance, power, and tackling risky missions. The level of autonomy given to these UGVs is a critical factor to consider. In many applications of multi-robotic systems like “search-and-rescue” missions, teamwork between human and robots is essential. In this paper, given a team of manned ground vehicles (MGVs) and unmanned ground vehicles (UGVs), the objective is to develop a model which can minimize the number of teams and total distance traveled while considering human-robot interaction (HRI) studies. The human costs of managing a team of UGVs by a manned ground vehicle (MGV) are based on human-robot interaction (HRI) studies. In this research, we introduce a combinatorial, multi objective ground vehicle path planning problem which takes human-robot interactions into consideration. The objective of the problem is to find: ideal number of teams of MGVs-UGVs that follow a leader-follower framework where a set of UGVs follow an MGV; and path for each team such that the missions are completed efficiently.

Real-time Trajectory Planning for Automated Vehicle Safety and Performance in Dynamic Environments

Safe trajectory planning for high-performance automated vehicles in an environment with both static and moving obstacles is a challenging problem. Part of the challenge is developing a formulation that can be solved in real-time while including the following set of specifications: minimum time-to-goal, a dynamic vehicle model, minimum control effort, both static and moving obstacle avoidance, simultaneous optimization of speed and steering, and a short execution horizon. This paper presents a nonlinear model predictive control-based trajectory planning formulation, tailored for a large, high-speed unmanned ground vehicle, that includes the above set of specifications. The ability to solve this formulation in real-time is evaluated using NLOptControl, an open-source, direct-collocation based, optimal control problem solver in conjunction with the KNITRO nonlinear programming problem solver. The formulation is tested with various sets of the specifications. A parametric study relating execution horizon and obstacle speed indicates that the moving obstacle avoidance specification is not needed for safety when the planner has a small execution horizon and the obstacles are moving slowly. However, a moving obstacle avoidance specification is needed when the obstacles are moving faster, and this specification improves the overall safety without, in most cases, increasing the solve-times. The results indicate that (i) safe trajectory planners for high-performance automated vehicles should include the entire set of specifications mentioned above, unless a static or low-speed environment permits a less comprehensive planner; and (ii) the resulting formulation can be solved in real-time.

Safe Learning Reference Governor: Theory and Application to Fuel Truck Rollover Avoidance

This paper proposes a learning reference governor (LRG) approach to enforce state and control constraints in systems for which an accurate model is unavailable; and this approach enables the reference governor to gradually improve command tracking performance through learning while enforcing the constraints during learning and after learning is completed. The learning can be performed either on a black-box type model of the system or directly on the hardware. After introducing the LRG algorithm and outlining its theoretical properties, this paper investigates LRG application to fuel truck (tank truck) rollover avoidance. Through simulations based on a fuel truck model that accounts for liquid fuel sloshing effects, we show that the proposed LRG can effectively protect fuel trucks from rollover accidents under various operating conditions.

A Repeated Commuting Driving Cycle Dataset with Application to Short-term Vehicle Velocity Forecasting

Vehicle velocity forecasting plays a critical role in operation scheduling of varying systems and devices for a passenger vehicle. The forecasted information serves as an indispensable prerequisite for vehicle energy management via predictive control algorithms or vehicle ecosystem control Co-design. This paper first generates a repeated urban driving cycle dataset at a fixed route in the Dallas area, aiming to simulate a daily commuting route and serves as a base for further energy management study. To explore the dynamic properties, these driving cycles are piecewise divided into cycle segments via intersection/stop identification. A vehicle velocity forecasting model pool is then developed for each segment, including the hidden Markov chain model, long short-term memory network, artificial neural network, support vector regression, and similarity methods. To further improve the forecasting performance, higher-level algorithms like localized model selection, ensemble approaches, and a combination of them are investigated and compared. Results show that (i) the segment-based forecast improves the forecasting accuracy by up to 20.1%, compared to the whole cycle-based forecast; and (ii) the hybrid localized model framework that combines dynamic model selection and an ensemble approach could further improve the accuracy by 9.7%. Moreover, the potential of leveraging the stopping location at an intersection to estimate the waiting time is also evaluated in this study.

Sensors in Autonomous Vehicles: A Survey

Research and technology in autonomous vehicles is beginning to become well recognized among computer scientists and engineers. Autonomous vehicles contain combination of GPS, LIDAR, cameras, RADAR and ultrasonic sensors (which are hardly ever included). These autonomous vehicles use no less than two sensing modalities, and usually have three or more. The goal of this research is to determine which sensor to use depending on the functionality of the autonomous vehicle and analyze the simi- larities and differences of sensor configurations (which may come from different industries too). This study summarizes sensors in four industries: personal vehicles, public transportation, smart farming, and logistics. In addition, the paper includes advantages and disadvantages of how each sensor configuration are helpful by taking into considerations the activity that has to be achieved in the autonomous vehicle. A table of results is incorporated to organize most of the sensors' availability in the market and their advantages and disadvantages. After comparing each sensor configuration, recommendations are going to be proposed for different scenarios in which some types of sensors will be more useful than others.

A Brief of the Economics of ADAS for Full Size Light Duty Pickup Trucks: Relating Effectiveness to Cost

Advanced Driver Assistance Systems (ADAS) have become increasingly common in vehicles in the last decade. The majority of studies have focused on smaller vehicles with gross vehicle weight rating (GVWR) under 5,000lbs, predominantly sedans, for their ADAS evaluations. While it is sensible to use this style of vehicle because it is ubiquitous worldwide for a typical vehicle body style, these studies neglect full-size light-duty pickup trucks, GVWR 5,000 – 10,000lbs, which are abundant on the roads in the United States. The increase in mass, higher center of gravity, and utilitarianism of the vehicles allows for unique conditions for studying the effects of ADAS. This work evaluates the effectiveness of ADAS in full-size light-duty pickup trucks across brands, representing 18% of registered vehicles in the US, at reducing severity of injury for occupants during accidents involving fatalities relative to expense of the ADAS technology. This work will illustrate the cost benefit of ADAS at reducing the severity of injuries for occupants of full-size light-duty pickup trucks for multiple different brands.

Can we Localize an AV from a Single Image? Deep-Geometric 6 DoF Localization in Topo-metric Maps

We describe a Deep-Geometric Localizer that is able to estimate the full six degrees-of-freedom (DoF) global pose of the camera from a single image in a previously mapped environment. Our map is a topo-metric one, with discrete topological nodes whose 6DOF poses are known. Each topo-node in our map also comprises of a set of points, whose 2D features and 3D locations are stored as part of the mapping process. For the mapping phase, we utilise a stereo camera and a regular stereo visual SLAM pipeline. During the localization phase, we take a single camera image, localize it to a topological node using Deep Learning, and use a geometric algorithm (PnP) on the matched 2D features (and their 3D positions in the topo map) to determine the full 6DOF globally consistent pose of the camera. Our method divorces the mapping and the localization algorithms and sensors (stereo and mono), and allows accurate 6DOF pose estimation in a previously mapped environment using a single camera. With results in simulated and real environments, our hybrid algorithm is particularly useful for autonomous vehicles (AVs) and shuttles that might repeatedly traverse the same route.

Operational Context Change Propagation Prediction on Autonomous Vehicles Architectures

Autonomous Vehicles (AV) are designed to operate in a specific Operational Context (OC), and the adaptability of the vehicle's architecture to its OC is considered a significant success criterion of the design. AV design projects are rarely started from scratch and are often based on reference architectures. As such, the reference architecture must be modified and adapted to the OC. The current literature on engineering change propagation does not provide a method to identify and anticipate the impact of OC changes on the AV reference architecture. This paper proposes a two-step method for OC change propagation: (1) Analyzing the direct impact of OC change and (2) evaluate the probabilities of indirect change propagation. The direct impact is assessed following a propagation path based upon a model mapping between an OC Ontology, operational situations, and Functional Chains. The effects of Functional Chain changes on the AV components are analyzed and evaluated by domain experts with Types of Changes and associated probabilities. A Bayesian Network is proposed to calculate the probabilities of indirect change propagation between component Types of Changes. The method's applicability and efficiency are validated on a real case design of AV architecture where the probabilities of the system components undergoing Types of Changes are evaluated.