Skip navigation

Welcome to the Boston Chapter of the IEEE Robotics and Automation Society. We invite you to join us and see renowned roboticists from both industry and academia in the Greater Boston area talk about their projects. We usually meet at the auditorium of Olin College in Needham, MA, on the second Tuesday of the month (except for June, July, and August), at 6 PM. Our meetings are free and open to the public, reservations are not required. Each meeting is followed by a dinner with the speaker at a local restaurant.

 

 

Next Meeting: Tuesday, December 10th 2013

Winter Social, SeminarĀ  and Chapter Elections

Boston Chapter, IEEE Robotics and Automation Society

Date: Tuesday, December 10th

 

Location:

MIT Lincoln Laboratory Beaver Works in Kendall Sq.
300 Technology Square
Suite 202 (second floor)
MIT Building NE45-202
Cambridge, MA 02139

Doors Open: 6PM
Refreshments / Light Dinner Served: 6:30PM
Elections: 6:45PM
Seminar Starts: 6:50PM
Social/Deserts: 7:30PM
Meeting Adjourns: 8:30PM

Seminar by Andrew Wang of MIT CSAIL

 

TITLE: Risk-driven Adjustable Autonomy for Disaster Response

ABSTRACT
Routing search-and-rescue vehicles during disaster response is risky due to unknown and dangerous conditions, such as flooded roads or underwater debris. During logistics support, evaluating this risk is crucial but difficult for a human operator. To reduce and focus the operator’s workload, we offer an adjustable autonomy approach. Our automation performs the bulk of assessing a logistics plan’s risk, but engages the operator where its assessment is weakest and would benefit most from human expertise.

First, we allow operators to specify upper bounds on acceptable risk by augmenting mission descriptions with chance constraints. Then, we augment a planning-and-execution architecture with a risk assessment capability. This capability includes an adaptive sampling component, which improves the risk assessment via relevant scouting. In the end, matching the assessed risk against the chance constraints decides the degree of interaction between automation and operator.

We demonstrate key components of our adjustable autonomy architecture on simulations of two disaster response scenarios: land rescue and underwater search. In the land rescue scenario, we assess the risk of path traversal by a ground logistics vehicle, and perform adaptive sampling via aerial scouts to identify safer paths. For underwater search, we assess temporal risk on an AUV’s battery life, and negotiate with the operator increasingly aggressive plan modifications to meet the chance constraints.