Grant: Discovery Grants Program - Individual

Multi--object manipulation is an umbrella term for any situation where a robot is interacting with multiple objects. Examples include forming clusters of objects, sorting them into groups, and arranging them into shapes. Usually, multi--object manipulation is studied using one robot with a gripper at the end of a manipulator's arm. This class of robots is very useful but expensive. Researchers who focus on single--robot manipulation have had great success recently. However, if we really want to develop robot systems that can make a positive impact on our lives and the economy, then we need to scale up to multi--robot systems. To scale up efficiently, the individual robots used should be inexpensive. Fortunately, there is an active research area based on the idea of a large group of relatively simple robots interacting to perform a task. This research area is swarm robotics. The proposed research program will address multi--object manipulation by adopting leaner strategies for robots to interact with objects. We will use "guidance fields", which are images that can be provided to a set of robots.  These images can be projected directly onto the robots' sensors, or referenced internally if the robots have the ability to localize (e.g. using a GPS-like system). A simple type of guidance field is a distance field, encoding the distance to a goal point throughout the space.  In our previous work, we showed that distance fields could be used for various tasks, including foraging and planar construction.  Another type of guidance field is a roadmap, encoding regions where robots are free to travel.  A roadmap can be used to keep robots from interfering with each other without them having to individually solve the computationally difficult problems of building their own maps and planning paths.  This is the general theme of the guidance field concept---provide locally relevant guidance to a swarm of robots to offload the computational burden from the robots themselves. Bringing multi--object manipulation capabilities to robot swarms could open up a wide range of applications in cleaning, recycling, manufacturing, and techniques for autonomous construction that could be applied in disaster zones or remote regions. This research has the potential to create new entrepreneurial opportunities. In Atlantic Canada there is a real need to increase regional efforts in automation, which will be key to a more diverse economic foundation. Our team will be spearheading this research into swarm--enabled automation to help Memorial University fulfill its important regional goal. This research also provides extremely fruitful training opportunities for HQP. Those participating in the research will learn a great deal from the practical problems of getting a swarm of robots to function while also staying abreast of state--of--the-art developments in the robotics research community.