Make way for the robots!
Problem being addressed
For various robotic applications such as search-and-rescue, terrain exploration, and environmental monitoring, an autonomous agent must gather information in uncertain environments with partial observability. In such settings, there is a trade-off between exploring the environment and exploiting the current belief about the environment to visit informative locations.
Adaptive Informative Path Planning with Multimodal Sensing that requires reasoning jointly about the effects of sensing and movement in terms of both energy expended and information gained. It uses an online solver to accommodate the state and observation spaces, which are exponentially large. For a tailored rollout policy to guide the online simulations, the researchers use an adaptive extension to a near-optimal algorithm for the fully observable variant of the problem.
Advantages of this solution
The resulting approach consistently outperforms the modified algorithm over multiple domains, and our rollout policy is a key contributor to this performance.
Solution originally applied in these industries
Electronic & Sensors Industry
Possible New Application of the Work
Robots are often used in search-and-rescue operations. In the aftermath of a disaster, survivors can be scattered uniformly across some terrain, and an aerial vehicle with limited energy can visit a number of locations and rescue however many people are stranded there (depending on the accessibility of a particular location to the survivors). Improving the path finding function will mean faster arrival and more people saved.
Mining robots can vary from robots that assist in excavation, to robotic devices with camera systems that detect gases and other materials. Most importantly, mining robots can be used to keep humans out of harm's way, and improved path finding algorithms can make their work more efficient.
Source URL: #############