Engineers at MIT have designed an autonomous robot that has socially aware navigation, allowing it to keep pace with foot traffic while observing the general codes of pedestrian conduct.
Most pedestrians follow certain social codes when navigating a hallway or a street: they keep to one side, pass on the other side, maintain a distance between themselves and others, and are ready to weave or change courses to avoid obstacles, while keeping up a steady walking pace.
The MIT robot resembles a knee-high kiosk on wheels, and in tests it was able to successfully avoid collisions while keeping up with the average flow of pedestrians.
The work was led by Yu Fan "Steven" Chen, a former MIT graduate student. According to Chen, socially aware navigation is a central capability required for mobile robots operating in environments that involve frequent interactions with pedestrians.
These include small robots that could be used in the future for package and food delivery, and personal mobility devices for transporting people in large crowded spaces.
The key to allowing a robot to make its way autonomously through a heavily trafficked environment is to solve four main challenges. It must know where it is in the world; it must be able to recognise its surroundings; it must be able to identify the optimal path to a given destination; and finally it must be able to physically execute its desired path.
For the first two challenges, the team outfitted the robot with off-the-shelf sensors such as webcams, a depth sensor, and a high-resolution lidar sensor. They used open-source algorithms to map the robot's environment and determine its position. Finally, they employed standard methods used to drive autonomous ground vehicles to control the robot.
The team needed to innovate on motion planning, however. Once the robot knew where it was in the world, and had determined how to follow trajectories, it needed to determine which trajectories to follow.
This is especially tricky in pedestrian-heavy environments, because individual paths can be difficult to predict. Some approaches require the robot to use its sensor data to predict the trajectory of all the pedestrians, and then compute an optimal path. But this can be slow, and in pedestrian traffic, the situation is always in flux.
Another approach is to have the robot use geometry or physics to quickly compute a path that avoids collisions. However, this too is problematic, because humans are unpredictable. Robots that use such approaches either collide with people or get pushed around by avoiding people. As a result, they do not fit into the socially accepted rules.
The team found a way around such limitations, enabling the robot to adapt to unpredictable pedestrian behaviour while continuously moving with the flow and following typical social codes of pedestrian conduct.
They used reinforcement learning, a type of machine learning approach, in which they performed computer simulations to train a robot to take certain paths, given the speed and trajectory of other objects in the environment.
The team also incorporated social norms into this offline training phase, in which they encouraged the robot in simulations to pass on the right, and penalised the robot when it passed on the left.
While the training can take a long time and a lot of computing power, it can also be done in simulation, with the robot taking what it learns from the simulation into the real world.
The researchers enabled the robot to assess its environment and adjust its path, every one-tenth of a second. In this way, the robot can continue rolling through a hallway at a typical walking speed of 1.2 metres per second, without pausing to reprogram its route.
This means the robot does not plan an entire path, but rather is constantly re-assessing and changing its velocity as required.
The team test-drove the robot in the busy, winding halls of an MIT building. The robot was able to drive autonomously for 20 minutes at a time. It rolled smoothly with the pedestrian flow, generally keeping to the right of hallways, occasionally passing people on the left, and avoiding any collisions.
[The knee-high robot attempts to fit in with unpredictable humans. Photo: MIT]