Integrating Time Performance in Global Path Planning for Autonomous Mobile Robots

Global path planners for autonomous mobile robots have been demonstrated to map optimal routes in structured environments. The search for the optimum path from a given starting point to a destination is usually formulated as a graph search problem. Over the years, much progress has been made in efficient methods for global motion planning in static environments. A number of both exact and approximate approaches in this field (see, for example, Latombe, 1991; Hwang & Ahuja, 1992; LaValle, 2006). However, real world scenarios are intrinsically dynamic due to the presence of non-modelled obstacles, both static and moving. Path planning in such environments is an area of active research. Different methods have also been proposed (Heero, 2006; Jaillet & Simeon, 2004). Some approaches use robot sensors to build a model of the environment incrementally even when the real world is static. In such circumstances, a new path must be calculated every time the model is updated. The online integration of this new information in a global map is usually complex and time consuming. Path planning with dynamic objects can be addressed by using explicit time representation to convert the problem into an equivalent static one that can then be solved using an existing static planner. However, this increases the dimensionality of the representation and requires exact motion models for moving objects (Szczerba & Chen, 1995; Hsu et al., 2002). Note that, in many circumstances (such as the presence of people near the robot in a tour-guide autonomous application), it is not possible to explicitly model the unpredictable behaviour of humans and robot movements with a time-extended representation (Philipsen et al., 2007). In contrast to previous works, our global path planning approach is able to take into account the effect of persistent delays caused by unpredictable objects (such as people in corridors or lobbies) by considering these delays for future path planner executions.