Design of a household robot with autonomous navigation for object detection and sorting. Credit: Bingjie Xu/Suzhou Industrial Park Institute of Vocational Technology, Qinglei Bu/Xi’an Jiaotong-Liverpool University.
Tackling the never-ending list of household chores might one day be the job of household robots, leaving people more free time for more important things in life.
A team of engineers in China developed a prototype service robot which can understand and respond to voice commands, navigate indoor environments and identify and sort various items.
Robots that make short work of vacuuming and mopping floors are already widely used. But sometimes what you really need is a hand tidying up.
“This robot is designed to be a comprehensive family assistant, capable of performing tasks that make daily life more convenient and enjoyable,” says Bingjie Xu from Suzhou Industrial Park Institute of Vocational Technology. “From picking up toys to fetching items, it can handle a wide range of objects with precision and care.”
The prototype is presented in a paper published in the journal Robot Learning.
Xu and collaborators designed and created the helper bot to have a lightweight, moveable chassis for ease of navigating homes. A depth camera, combined with an integrated object recognition algorithm, allows the robot to detect target objects in real-time.
Experiments revealed that its 3D-printed robotic arm and soft, flexible gripper can grab hold of objects of different sizes and weights, including: an orange, bowl, bottle, stuffed bear, book, umbrella, handbag and potted plant.
To generate a 2-dimensional map of its environment, the robot combines a mapping algorithm with LiDAR – a method which measures the distance to an object by shining a laser and determining the time taken for the reflected light to return.
“The robot can move seamlessly through rooms, avoid obstacles, and reach designated locations on its own,” says Yangzesheng Lu of Xi’an Jiaotong-Liverpool University. “This feature is crucial for tasks like picking up toys scattered around the house or fetching items from different rooms.”
The system is also equipped with a speech recognition module which allows it to process simple voice commands such as “turn left” or “follow”.
Experiment participants were instructed to give a series of voice commands to the robot under different conditions which, according to the authors, “simulated daily home scenarios, including different areas such as the living room and kitchen, and considered factors such as varying background noise and speaking distances”.
“Upon receiving the voice commands, the robot processed them through its built-in speech recognition system and provided feedback to the participants. The participants then judged the accuracy of the recognition,” they write.
“In a quiet environment, the system’s average recognition accuracy is 83.8%, indicating that 83.8% of the recognised samples are correct. In environments with background noise, the accuracy drops slightly but remains above 75%.”
The researchers suggest intelligent household robots are poised to become an indispensable part of human life.
“We are excited about the potential applications of this robot in everyday life,” says co-author Dr Qinglei Bu, also from Jiaotong-Liverpool University.
“Future work will focus on enhancing the robot’s object detection accuracy and integrating large language models to improve its semantic understanding capabilities.”
The authors acknowledge there remains a gap between the current version of their prototype and mass production.
“Due to the inherent complexity of the system and the potential risks, full-scale testing in real-home environments has not yet been undertaken,” they write. “We recognise that such real-world testing will be critical in refining and perfecting the robot’s various functions.”