Reactive approach to on-line path planning for robot manipulators in dynamic environments* Margarita Mediavilla, José Luis González, Juan Carlos Fraile and José Ramón Perán Departamento de Ingeniería de Sistemas y Automática, E.T.S.I.I., Universidad de Valladolid, Valladolid (Spain) E-mail: marga@eis.uva.es (Received in Final Form: February 1, 2002) SUMMARY This paper describes a new approach to path planning of robot manipulators with many degrees of freedom. It is designed for on-line motion in dynamic and unpredictable environments. The robots react to moving obstacles using a local and reactive algorithm restricted to a subset of its configuration space. The lack of a long-term view of local algorithms (local minima problems) is solved using an off- line pre-planning stage that chooses the subset of the configuration space that minimises the probability of not finding collision free paths. The approach is implemented and tested on a system of three Scorbot-er IX five link robots. KEYWORDS: Path planning; Dynamic environments; On-line motion; Scorbot-er IX robots 1. INTRODUCTION It is well known that motion planning for robot manip- ulators with many degrees of freedom is a complex task. This is probably the reason why the research on this area has been mostly restricted to static environments, where the obstacles are still or at least follow known trajectories. 1–4 Since most path planning is designed for static environ- ments, robot manipulators (in industry as well as in many research projects) are restricted to pre-calculated trajectories and subject to rigid timings. These features make them too rigid and unable to adapt to changes. Increasing the reactivity of robot motion would have important advantages, If the robots had reactive path planning algorithms they could: • react to unexpected events such as moving obstacles or faults, • move without a previous model of their environment, based only on sensor information, • be less dependent on pre-calculated trajectories, since path planning could be done on-line. In the last four years, our research group has been working on a prototype of a multi-manipulator system 5–8 located in our laboratories of the University of Valladolid*. Our multi manipulator system is designed as an assembling cell that achieves a high production rate by removing the rigid timings and delays from the system. The three robots of our system share a good portion of their working space, but the path planning method that we are developing enables them to operate despite the fact that other robots might also be moving in the same area. The high production rate is achieved by making the robots depend only on the rate of arrival of the supplies. This means that, when a supply comes in, one of the robots starts its motion towards it, irrespective of whether the rest of the robots are at that time already in motion or not. This kind of behaviour needs the robots to have an on-line and reactive planning method. Path planning approaches for robotic manipulators can broadly be categorized into the two classes of global and local methods: Global methods 9–12 are computationally very expensive, and computational cost increases rapidly as a function of the number of manipulator joints. Furthermore, they are not applicable when the obstacles are unmodeled or subject to uncertainties. Local methods, on the other hand, 13–15 are reactive and can be used for real-time path planning, but most of them are very limited in their capabilities and easily get trapped in local minima. It is the problem of the local minima that has avoided potential field methods from becoming a valid reactive path planning framework for manipulators. This paper presents a new approach to reactive on-line path planning for robot manipulators. It is designed for robots that move in a dynamic environment subject to uncertainties. By dynamic environment subject to uncertain- ties we understand a workplace with still and moving obstacles where the position of the obstacles at each instant of time is known, but where the entire trajectory of the obstacles cannot, because the position is either influenced by external perturbations, or conditioned by our own path planning. There are not many approaches to on-line motion planning of robot manipulators in the literature. Li and Latombe 16 Ek describe an application of on-line path plannning of two Scara robots. They use a non-reactive method based on global planning. They obtain on-line path * Part of this work was carried out while the first author was visiting Prof. Gupta at the robotics laboratory, Simon Fraser University, Canada. * This work has been supported by the Comisión Interministerial de Ciencia y Tecnología. CICYT, TAP95/0092. Robotica (2002) volume 20, pp. 375–384. © 2002 Cambridge University Press DOI: 10.1017/S0263574702004071 Printed in the United Kingdom