Path Planning for Information Gathering with Lethal Hazards and No Communication.
WAFR(2018)
摘要
We consider a scenario where agents search for targets in a hazardous environment that prevents communication. Agents in the field cannot communicate, and hazards are only directly observable by the agents that are destroyed by them. Thus, beliefs about hazard locations must be inferred by sending agents to travel along various paths and then observing which agents survive. In other words, agent survival along a path can be used as a sensor for hazard detection; we call this form of sensor a “path-based sensor”. We present a recursive Bayesian update for path-based sensors, and leverage it to calculate the expected information gained about both hazards and targets along a particular path. We formalize the resulting iterative information based path planning problem that results from this scenario, and present an algorithm to solve it. Agents iteratively foray into the field. The next path each agent follows is calculated to maximize a weighted combination of the expected information gained about targets and hazards (where the weighting is defined by user preferences). The method is evaluated in Monte Carlo simulations, and we observe that it outperforms other techniques.
更多查看译文
关键词
path planning,information gathering,lethal hazards
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络