Agile Robotics for Logistics

The Agile Robotics for Logistics effort (2008-2011) was a project at the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT that developed a voice-commandable robotic forklift for the purposes of improving the efficiency and safety of logistics operations in unstructured and semi-structured outdoor environments, such as those associated with disaster relief and the military supply chain. Led by the late Seth Teller, the project was a collaboration between MIT CSAIL, Draper Laboratory, MIT Lincoln Laboratory, and the U.S. Army Combined Arms Support Command (CASCOM) and Logistics Innovation Agency (LIA). The project was supported by the Department of Defense: Research and Engineering (DDRE), CASCOM, and LIA.

The voice-commandable autonomous forklift executed commands conveyed via natural language speech or stylus-based gestures to approach, engage, transport and place palletized cargo in minimally-structured outdoor settings. As part of the project, the team developed and integrated capabilities that allowed the robot to operate effectively alongside people within existing unstructured environments, such as military Supply Support Activities (SSAs, outdoor warehouses). The robot was designed to operate safely outdoors on uneven terrain and in unprepared environments, without the need for specially-placed fiducial markers, guidewires or other localization infrastructure, or specialized pallets or trucks, directly alongside people on foot and in human-driven vehicles. Fundamental to the design, people were able to effectively command and interact with the robot with little-to-no prior training. The robot was designed such that its operation was acceptable to existing (military) personnel and was consistent with their current operational practices and culture. The system exhibited several novel characteristics that were fundamental to its ability to operate safely and effectively despite challenging operational requirements:

• autonomous operation in dynamic, minimally-prepared, real-world environments, outdoors on uneven terrain without reliance on precision GPS, and in close proximity to people;
• speech understanding in noisy environments;
• indication of robot state and imminent actions to bystanders;
• persistent visual memories of objects in the environment;
• multimodal interaction that includes natural language speech and pen-based gestures grounded in a world model common to humans and the robot; and
• robust, closed-loop pallet manipulation using only local sensing.

For a self-contained summary of the Agile Robotics for Logistics effort, please see our IJRR 2015 paper.

Questions: If you have any questions, or would like more information about the project, please contact Matthew Walter.

Publications