The ratio of guide dogs to blind people in China is roughly one service pooch for every 40,000 people. Researchers are currently developing a six-legged, AI-enhanced robodog to close the gap and safely guide users to their destinations.
The press release on the project states that, according to the China Blind Association, there are more than 17 million blind people in China, yet only just over 400 active guide dogs. The guide robot being tested by Professor Feng Gao and team from Shanghai Jiao Tong University's School of Mechanical Engineering was developed to fill the need gap.
The guide robot won't need feeding, save for regular plug-ins to charge its battery, and won't need costly individual training or a regular supply of doggy treats. It's got six legs rather than the four of its biological cousins, which Lenovo recently claimed gave its Daystar Bot GS model "unmatched stability and maneuverability." And it hauls a suite of sensors on its back that includes depth cameras and radar fed into machine learning algorithms to help it get around.
The bot can generate 3D environment maps and position itself within the virtual locale, autonomously navigate the busy streets, plan paths and routes, dynamically avoid obstacles and even detect signals at pedestrian crossings. It can receive voice commands from the user thanks to "deep-learning end-to-end speech recognition models," but has also been treated to tactile and force feedback interaction modes. That last one means that it can be instructed to slow down by pulling back on the telescoping pole mounted to its back, or to speed up by pushing this "leash."
The top speed is reported to be 3 meters per second, but the normal range out of the robo-kennel is between 0.6 and 0.8 m/s. Though it can operate without needing to get online, being connected will allow for expanded use within a home environment, it could also be relied on in an emergency and "lead visually impaired individuals to more places."
Gao's engineering team is currently field testing the robo guide dog, and is working with Suochen Robot Co towards commercialization. There's no word on when they'll go into production, but the price tag per pooch is not likely to be wallet friendly. The video below has more.
Source: Shanghai Jiao Tong University
My concern is that the sensor/compute suite (rather like some "self-driving" cars) will have situations where it doesn't recognize oncoming dangers, and will lead users to injury. And if one guide robot has these loopholes, they all will.