Scott Pendleton

Autonomous golf cars for public trial of mobility-on-demand service

We detail the design of autonomous golf cars which were used in public trials in Singapores Chinese and Japanese Gardens, for the purpose of raising public awareness and gaining user acceptance of autonomous vehicles. The golf cars were designed to be robust, reliable, and safe, while operating under prolonged durations. Considerations that went in to the overall system design included the fact that any member of the public had to not only be able to easily use the system, but to also not have the option to use the system in an unintended manner. This paper details the hardware and software components of the golf cars with these considerations, and also how the booking system and mission planner facilitated users to book for a golf car from any of ten stations within the gardens. We show that the vehicles performed robustly throughout the prolonged operations with a small localization variance, and that users were very receptive from the user survey results.

Autonomous vehicle planning system design under perception limitation in pedestrian environment

Autonomous driving within the pedestrian environment is always challenging, as the perception ability is limited by the crowdedness and the planning process is constrained by the complicated human behaviors. In this paper, we present a vehicle planning system for self-driving with limited perception in the pedestrian environment. Acknowledging the difficulty of obstacle detection and tracking within the crowded pedestrian environment, only the raw LIDAR sensing data is employed for the purpose of traversability analysis and vehicle planning. The designed vehicle planning system has been experimentally validated to be robust and safe within the populated pedestrian environment.

Scalable Cooperative Localization with Minimal Sensor Configuration

Localization of distributed robots can be improved by fusing the sensor data from each robot collectively in the network. This may allow for each individual robot’s sensor configuration to be reduced while maintaining an acceptable level of uncertainty. However, the scalability of a reduced sensor configuration should be carefully considered lest the propagated error become unbounded in large networks of robots. In this paper, we propose a minimal but scalable sensor configuration for a fleet of vehicles localizing on the urban road. The cooperative localization is proven to be scalable if the sensors’ data are informative enough. The experimental results justify that pose uncertainty will remain at an acceptable level when the number of robots increases.

Teleoperation of On-Road Vehicles via Immersive Telepresence Using Off-the-shelf Components

The quality of visual information and response time are crucial aspects of any modern teleoperation system. This is especially true for operation of on-road vehicles, which must function in highly dynamic, unforgiving environments. In this work we demonstrate that a suitable teleoperation system can be exclusively composed of low-cost off-the-shelf components yet still meet the high performance demands of remotely driving a car on the road. The user is given immersive situational awareness through an on-board head-mounted display linked to an actuated stereoscopic camera, thereby maintaining depth perception and intuitive camera control. Communication speeds are evaluated over various wireless connection types, and a usability study shows that the system allows for advanced driving maneuvers while remotely controlled. 3G and 4G data networks are demonstrated to provide adequate bandwidth for the task given proper data compression, thus expanding the potential range for teleoperation. Applications for such a system are further discussed, extending to fleet management and autonomous vehicle safety measures.

Numerical Approach to Reachability-Guided Sampling-Based Motion Planning Under Differential Constraints

This paper presents a new method for motion planning under differential constraints by incorporating a numerically solved discretized representation of reachable state space for faster state sampling and nearest neighbor searching. The reachable state space is solved for offline and stored into a “reachable map” which can be efficiently applied in online planning. State sampling is performed only over states encompassed by the reachable map to reduce the number of unsuccessful motion validity checking queries. The nearest neighbor distance function is revised such that only reachable states are considered, with states which are unreachable or only reachable beyond a designated time horizon disregarded. This method is generalized for application to any control system, and thus can be used for vehicle models where analytical solutions cannot be found. Greater improvement is expected for more constrained systems where motion checking cost is relatively high. Simulation results are discussed for case studies on a holonomic model and a Dubins car model, both with maximum speed limitation and time included as a dimension in the configuration space, where planning speed (measured by tree growth rate) can be improved through reachability guidance in each system by at least a factor of 3 and 9, respectively.

A parallel autonomy research platform

We present the development of a full-scale “parallel autonomy” research platform including software and hardware. In the parallel autonomy paradigm, the control of the vehicle is shared; the human is still in control of the vehicle, but the autonomy system is always running in the background to prevent accidents. Our holistic approach includes: (1) a drive-by-wire conversion method only based on reverse engineering mounting of relatively inexpensive sensors onto the vehicle implementation of a localization and mapping system, (4) obstacle detection and (5) a shared controller as well as (6) integration with an advanced autonomy simulation system (Drake) for rapid development and testing. The system can operate in three modes: (a) manual driving, (b) full autonomy, where the system is in complete control of the vehicle and (c) parallel autonomy, where the shared controller is implemented. We present results from extensive testing of a full-scale vehicle on closed tracks that demonstrate these capabilities.

Pedestrian Notification Methods in Autonomous Vehicles for Multi-Class Mobility-on-Demand Service

In this paper, we describe methods of conveying perception information and motion intention from self driving vehicles to the surrounding environment. One method is by equipping autonomous vehicles with Light-Emitting Diode (LED) strips to convey perception information; typical pedestrian-driver acknowledgement is replaced by visual feedback via lights which change color to signal the presence of obstacles in the surrounding environment. Another method is by broadcasting audio cues of the vehicles motion intention to the environment. The performance of the autonomous vehicles as social robots is improved by building trust and engagement with interacting pedestrians. The software and hardware systems are detailed, and a video demonstrates the working system in real application. Further extension of the work for multi-class mobility in human environments is discussed.

Situation-aware decision making for autonomous driving on urban road using online POMDP

As autonomous vehicles begin venturing on the urban road, rational decision making is essential for driving safety and efficiency. This paper presents a situation-aware decision making algorithm for autonomous driving on urban road. Specifically, an urban road situation model is proposed first for proper environment representation, thereafter the situation-aware decision making problem is modeled as a Partially Observable Markov Decision Process (POMDP) and solved in an online manner. The proposed algorithm has been extensively evaluated, which is general enough for autonomous driving in various urban road scenarios, including leader following, collision avoidance and traffic negotiation at both T-junction and roundabout.

Autonomous personal mobility scooter for multi-class mobility-on-demand service

In this paper, we describe the design and development of an autonomous personal mobility scooter that was used in public trials during the 2016 MIT Open House, for the purpose of raising public awareness and interest about autonomous vehicles. The scooter is intended to work cooperatively with other classes of autonomous vehicles such as road cars and golf cars to improve the efficacy of mobility-on-demand transportation solutions. The scooter is designed to be robust, reliable, and safe, while operating under prolonged durations. The flexibility in fleet expansion is shown by replicating the system architecture and sensor package that has been previously implemented in a road car and golf cars. We show that the vehicle performed robustly with small localization variance. A survey of the users shows that the public is very receptive to the concept of the autonomous personal mobility device.

Multi-class autonomous vehicles for mobility-on-demand service

Mobility-on-Demand (MoD) services can be enhanced through use of Autonomous Vehicles (AVs) to reduce manpower costs (among other benefits), and use of multiple classes of vehicles to expand service coverage and accessibility. This work presents a functional proof of concept MoD system accessible via mobile phone to utilize three classes of vehicles in combination: a road car, buggy, and mobility scooter. A common software architecture and primary sensor suite allows for flexible replication to additional vehicles regardless of vehicle model or even class type. Benefits of using these three classes in a combined service are discussed, and details are provided concerning the unique aspects of the conversion and systems integration for each vehicle. Various safety features are implemented to ensure safe user interaction with all AVs. The complete MoD system is tested in uncontrolled pedestrian environments as well as on road with real vehicular traffic.