Most of the existing coverage path planning (CPP) algorithms do not have the capability of enabling a robot to handle unexpected changes in the coverage area of interest. Examples of unexpected changes include the sudden introduction of stationary or dynamic obstacles in the environment and change in the reachable area for coverage (e.g., due to imperfect base localization by an industrial robot). Thus, a novel adaptive CPP approach is developed that is efficient to respond to changes in real-time while aiming to achieve complete coverage with minimal cost. As part of the approach, a total reward function that incorporates three rewards is designed where the first reward is inspired by the predator–prey relation, the second reward is related to continuing motion in a straight direction, and the third reward is related to covering the boundary. The total reward function acts as a heuristic to guide the robot at each step. For a given map of an environment, model parameters are first tuned offline to minimize the path length while assuming no obstacles. It is shown that applying these learned parameters during real-time adaptive planning in the presence of obstacles will still result in a coverage path with a length close to the optimized path length. Many case studies with various scenarios are presented to validate the approach and to perform numerous comparisons.

中文翻译：

---

PPCPP：一种基于 Predator-Prey 的自适应覆盖路径规划方法

大多数现有的覆盖路径规划 (CPP) 算法不具有使机器人能够处理感兴趣的覆盖区域中的意外变化的能力。意外变化的例子包括环境中突然引入静止或动态障碍物以及可覆盖区域的变化（例如，由于工业机器人的基地定位不完善）。因此，开发了一种新颖的自适应 CPP 方法，它可以有效地实时响应变化，同时旨在以最小的成本实现完全覆盖。作为该方法的一部分，设计了包含三个奖励的总奖励函数，其中第一个奖励受到捕食者-猎物关系的启发，第二个奖励与沿直线方向的持续运动有关，第三个奖励与覆盖有关边界。总奖励函数作为启发式方法在每一步引导机器人。对于给定的环境地图，首先离线调整模型参数以在假设没有障碍的情况下最小化路径长度。结果表明，在存在障碍物的情况下，在实时自适应规划期间应用这些学习参数仍然会导致覆盖路径的长度接近优化路径长度。提供了许多具有各种场景的案例研究，以验证该方法并进行大量比较。结果表明，在存在障碍物的情况下，在实时自适应规划期间应用这些学习参数仍然会导致覆盖路径的长度接近优化路径长度。提供了许多具有各种场景的案例研究，以验证该方法并进行大量比较。结果表明，在存在障碍物的情况下，在实时自适应规划期间应用这些学习参数仍然会导致覆盖路径的长度接近优化路径长度。提供了许多具有各种场景的案例研究，以验证该方法并进行大量比较。