Michal Čáp

A Survey of Motion Planning and Control Techniques for Self-Driving Urban Vehicles

Self-driving vehicles are a maturing technology with the potential to reshape mobility by enhancing the safety, accessibility, efficiency, and convenience of automotive transportation. Safety-critical tasks that must be executed by a self-driving vehicle include planning of motions through a dynamic environment shared with other vehicles and pedestrians, and their robust executions via feedback control. The objective of this paper is to survey the current state of the art on planning and control algorithms with particular regard to the urban setting. A selection of proposed techniques is reviewed along with a discussion of their effectiveness. The surveyed approaches differ in the vehicle mobility model used, in assumptions on the structure of the environment, and in computational requirements. The side by side comparison presented in this survey helps to gain insight into the strengths and limitations of the reviewed approaches and assists with system level design choices.

Prioritized Planning Algorithms for Trajectory Coordination of Multiple Mobile Robots

In autonomous multirobot systems one of the concerns is how to prevent collisions between the individual robots. One approach to this problem involves finding coordinated trajectories from start to destination for all the robots and then letting the robots follow the preplanned coordinated trajectories. A widely used practical method for finding such coordinated trajectories is “classical” prioritized planning, where robots plan sequentially one after another. This method has been shown to be effective in practice, but it is incomplete (i.e., there are solvable problem instances that the algorithm fails to solve) and it has not yet been formally analyzed under what circumstances is the method guaranteed to succeed. Further, prioritized planning is a centralized algorithm, which makes the method unsuitable for decentralized multirobot systems. The contributions of this paper are: a) an adapted version of classical prioritized planning called revised prioritized planning with a formal characterization of a class of instances that are provably solvable by this algorithm and b) an asynchronous decentralized variant of both classical and revised prioritized planning together with a formal analysis showing that the algorithm terminates and inherits completeness properties from its centralized counterpart. The experimental evaluation performed in simulation on realworld indoor maps shows that: a) the revised version of prioritized planning reliably solves a wide class of instances on which both classical prioritized planning and popular reactive technique ORCA fail and b) the asynchronous decentralized implementation of classical and revised prioritized planning finds solution in large multirobot teams up to 2x-faster than the previously proposed synchronized decentralized approach. Note to Practitioners-Consider a large warehouse in which the goods are stored and retrieved by autonomous mobile robots. One way to deal with possible collisions between the robots is to ignore interactions between the vehicles during the route planning for each robot and handle the conflicts only during the route execution. However, such an approach is prone to deadlocks, i.e., to situations during which some of the robots mutually block each other, cannot proceed and fail to complete their transportation task. An alternative approach would involve planning collision-free routes for each robot before the robots start executing them. However, the current methods that guarantee ability to find a solution to any such coordination problem are not applicable in practice due to their high computational complexity. Instead, a simple and computationally efficient approach in which robots plan their routes sequentially one after another (classical prioritized planning) is often used for finding coordinated trajectories even though the algorithm is known to fail on many dense problem instances. In this paper, we show that a simple adaptation of this classical algorithm called revised prioritized planning is guaranteed to find collision-free trajectories for a well-defined class of practical problems. In particular, if the system resembles human-made transport infrastructures by requiring that the start and destination position of each vehicle must never obstruct other vehicles from moving, then the proposed approach is guaranteed to provide a solution. For instance, in our warehouse multirobot system example, the collision-free routes can be efficiently computed by the revised prioritized planning approach. This paper formally characterizes the problem instances for which the method is guaranteed to succeed. Further, we propose a new asynchronous decentralized adaptation of both classical and revised prioritized algorithm that can be used in multirobot systems without a central solver. This technique can be used to find coordinated trajectories just by running a simple asynchronous negotiation protocol between the individual robots. This paper provides an analysis showing that the asynchronous decentralized implementations of classical and revised prioritized planning exhibit desirable theoretical properties and an experimental comparison of performance of different variations of centralized and decentralized prioritized planning algorithms.

Mixed-reality testbeds for incremental development of HART applications

An incremental process for developing human-agent-robot applications uses mixed-reality testbeds of varying fidelity and size.

Developing Multiagent Algorithms for Tactical Missions Using Simulation

The development process and simulation architecture presented here help narrow the gap between how theoretical AI algorithms are traditionally designed and validated and how practical algorithms for controlling robotic assets in simulated tactical missions are developed.

Asynchronous decentralized prioritized planning for coordination in multi-robot system

In this paper, the multi-robot motion coordination planning problem is addressed. Although a centralized prioritized planning strategy can be used to solve the problem, we rather consider a decentralized variant, which is a more suitable for a robotic system of cooperating unmanned aerial vehicles (UAVs) due to communication limitations, privacy concerns, and a better exploitation of computational resources distributed among the individual robots. However, the existing decentralized prioritized planning algorithm contains synchronization points that all the agents must be able to pass synchronously, which is impractical and inefficient for a real-world deployment of the robotic systems. Therefore, we introduce a new asynchronous decentralized prioritized planning algorithm and show that the method can converge faster than both the synchronous decentralized and centralized algorithms. Further, we demonstrate the applicability of the proposed method as a coordination mechanism within a complex mission planning for a real robotic system consisting of several autonomous UAVs.

Belief/goal sharing modules for BDI languages

This paper proposes a modularisation framework for BDI based agent programming languages developed from a software engineering perspective. Like other proposals, BDI modules are seen as encapsulations of cognitive components. However, unlike other approaches, modules are here instantiated and manipulated in a similar fashion as objects in object orientation. In particular, an agents mental state is formed dynamically by instantiating and activating BDI modules. The agent deliberates on its active module instances, which interact by sharing their beliefs and goals. The formal semantics of the framework are provided and some desirable properties of the framework are shown.

Towards Incremental Development of Human-Agent-Robot Applications using Mixed-Reality Testbeds

Testing is essential part of the development of human-agent-robot team (HART) applications. Individual algorithms in such applications cannot be tested in isolation as their performance depends significantly on complex interactions among distributed software code, humans, hardware and the target environment. Any testing involving robots and human actors is, however, time-consuming and costly. We therefore propose an incremental development framework employing mixed-reality testbeds, which can reduce testing cost by replacing parts of the application and surrounding reality with synthetic computational models. The proposed framework introduces the concept of testbed fidelity and proposes how test reliability and cost should be managed to maximize the effectiveness of the development process. The framework is illustrated on two example applications in the domain of multi-UAV tracking and anti-maritime piracy operations.

Duckietown: An open, inexpensive and flexible platform for autonomy education and research

Duckietown is an open, inexpensive and flexible platform for autonomy education and research. The platform comprises small autonomous vehicles (“Duckiebots”) built from off-the-shelf components, and cities (“Duckietowns”) complete with roads, signage, traffic lights, obstacles, and citizens (duckies) in need of transportation. The Duckietown platform offers a wide range of functionalities at a low cost. Duckiebots sense the world with only one monocular camera and perform all processing onboard with a Raspberry Pi 2, yet are able to: follow lanes while avoiding obstacles, pedestrians (duckies) and other Duckiebots, localize within a global map, navigate a city, and coordinate with other Duckiebots to avoid collisions. Duckietown is a useful tool since educators and researchers can save money and time by not having to develop all of the necessary supporting infrastructure and capabilities. All materials are available as open source, and the hope is that others in the community will adopt the platform for education and research.

Provably safe and deadlock-free execution of multi-robot plans under delaying disturbances

One of the standing challenges in multi-robot systems is the ability to reliably coordinate motions of multiple robots in environments where the robots are subject to disturbances. We consider disturbances that force the robot to temporarily stop and delay its advancement along its planned trajectory which can be used to model, e.g., passing-by humans for whom the robots have to yield. Although reactive collision-avoidance methods are often used in this context, they may lead to deadlocks between robots. We design a multi-robot control strategy for executing coordinated trajectories computed by a multi-robot trajectory planner and give a proof that the strategy is safe and deadlock-free even when robots are subject to delaying disturbances. Our simulations show that the proposed strategy scales significantly better with the intensity of disturbances than the naive liveness-preserving approach. The empirical results further confirm that the proposed approach is more reliable and also more efficient than state-of-the-art reactive techniques.

Using agent technology to build a real-world training application

Using staff personnel for playing roles in simulation-based training (e.g. team mates, adversaries) elevates costs, and imposes organizational constraints on delivery of training. One solution to this problem is to use intelligent software agents that play the required roles autonomously. BDI modeling is considered fruitful for developing such agents, but have been investigated typically in toy-worlds only. We present the use of BDI agents in training a complex real-world task: on-board fire fighting. In a desktop simulation, the trainee controls the virtual character of the commanding officer. BDI-agents are developed to generate the behavior of all other officers involved. Additionally, agents are implemented to manage the information flow between the agents and the simulation, to control the scenario, and to tutor the trainee. In this paper we describe the design of the application, the functional and technical requirements, and our experiences during implementation.