MIT researchers develop a robot to help users get dressed

There are many activities most of us take for granted that we perform every day in our daily lives. Tasks such as getting dressed are easy for most people, but for those with various disabilities, getting dressed can be a major challenge. Researchers around the world are working to develop robots that can assist people with the activities of daily living, such as getting dressed and eating.Researchers from the MIT CSAIL laboratory are developing robots to work with humans to help with things such as getting dressed. The challenge is that robots lack the hardwired behavioral awareness and control humans have, so getting them to collaborate well with people requires planning and coordination. For example, a human assistant can do things like pour a cup of coffee without spilling any, but a robot can be challenged with that simple task. MIT researchers have created a new algorithm to help a robot find efficient motion plans to ensure the physical safety of any human it's working with.

The robot the team designed helps people put on a jacket, which is a simple task but could be useful for helping users with disabilities or limited mobility. The algorithm was developed to help prevent physical harm to a person without unnecessarily impacting the efficiency of the task the robot is performing. The algorithm allows the robot to make non-harmful impact with humans so the robot can find efficient trajectories to address the human with a safety guarantee.

To make the algorithm work required proper human modeling in so far as how the human moves, reacts and responds. Human modeling is required to enable successful robot motion planning in a human-robot interactive task. Robots can interact fluently if the human model is perfect, but the problem is there's not a flawless blueprint. Rather than having a single default model where the robot only understands one potential reaction, the team could give the machine an understanding of multiple potential models.

This allows the robot to closely mimic how humans can understand other humans. The robot is also capable of gathering more data to reduce its uncertainty and refine its own models. The team knows that impacts with a person the robot is trying to help can't be avoided, so they allow it to make non-harmful contact allowing the robot to dress the person in a shorter period.