From Berkeley University, the first four-legged robotic guide dog

While there are already several examples of robotic guide dogs , most of them use wheel-based locomotion systems and have stiff sticks as leashes. These characteristics do not make these driving systems very versatile, especially in the vicinity of narrow and crowded spaces. The robotic guide dog developed by researchers at Berkeley University, on the other hand, has the appearance of a real dog because it uses all four legs to move.

The idea of ​​the “hybrid model” for the robotic guide dog

At the base of the robot created by the researchers of the Hybrid Robotics Group is the concept of hybrid model. The system as a whole consists, in fact, of three fundamental parts: the human to be guided, the dog-robot and the leash. The idea of ​​the researchers was precisely that of wanting to artificially replicate the interaction that is created between man and dog when he pulls the leash. Thanks to a force sensor positioned at the base of the leash, scientists have developed a guiding system in which the interaction between the robot and the human being is based on the tension exerted by the robotic dog on the leash itself .

Photograph of the robotic guide dog by UC Berkeley University researchers. Source: TechXplore.
Photograph of the robotic guide dog by UC Berkeley University researchers. Source: TechXplore.

For example, when the robot-leash-human system enters a tight space, such as a corner of a corridor, the robot can first loosen the leash. This allows the robot to reposition and reorient itself without directly affecting the position of the human being. Once repositioned, the robot can pull the leash again to guide the human correctly through the narrow region.

Zhongyu Li, a researcher on the team.

How the robotic guide dog works

One of the fundamental characteristics of mobile robots is to be able to map the entire surrounding environment. The quadruped of the Hybrid Robotics Group is equipped with sensors that allow it to observe and map all the space in which it is located . Thanks to a 2D LIDAR sensor and an RGB camera , it is able to build a real map (identifying the obstacles present) and to estimate both its own position and that of the person inside it. The camera, in fact, is mounted on a pedestal able to rotate and tilt just like a head does.

In the image you can see all the details of the guidance system consisting of the human, the robot (Mini Cheetah) and the leash. Furthermore, the figure also shows the force sensor, used by the leash to monitor the tension exerted on it by the dog, and all the other sensors used by the robot to orient itself in space. Source: Hybrid Robotics Group, UC Berkeley University.
In the image you can see all the details of the guidance system consisting of the human, the robot (Mini Cheetah) and the leash. Furthermore, the figure also shows the force sensor, used by the leash to monitor the tension exerted on it by the dog, and all the other sensors used by the robot to orient itself in space. Source: Hybrid Robotics Group, UC Berkeley University.

Once you have estimated your position and that of the human within the map, the robot defines a trajectory to follow (avoiding obstacles) trying to minimize the time taken to reach the arrival point, the tension with which the leash is pulled and the speed with which the quadruped moves forward.

The robot-leash-human model

To describe the entire driving system, the team researchers defined a dynamic model capable of representing the robot-leash-human system in both the tight and loose leash conditions. In the first case they started from the hypothesis that the blind individual is guided by the dog moving in the same direction as the force applied by the robot on the leash. In this way, the person feels the force only when the leash is taut. When the leash is loose, however, it is only the robot that is free to move in space. In both situations, the equations of the dynamic model were obtained simply from geometric relations.

The results of the experiments

The experiments conducted in the laboratories of Berkeley University have shown some very promising results . The robotic guide dog, tested within a narrow and obstacle-filled course, was able to correctly and safely guide a blindfolded person in three different circumstances. In addition, the researchers noted an extremely cautious behavior on the part of the quadruped : near the second narrow passage, as can also be seen in the video, the robot tends to remain in the configuration with the leash loose, until it realizes that being able to make his master continue through the narrow passage without problems.

The team argues that any future developments of the robotic guide dog may involve modeling more complex human behaviors . Some of them, for example, provide for the traction of the leash also by the man, making the system even more realistic. The encouraging results of the experiments pave the way for a new era of service robotics, in which robots will increasingly be used for purposes of assisting frail or disabled people.

Article by Augusto Bozza

The article From Berkeley University the first four-legged robotic guide dog comes from Tech CuE | Close-up Engineering .