The traditional manual harvesting process remains labor-intensive, leading to increased costs in agricultural operations. In parallel, fully automated robots face challenges in precision when navigating complex environments. Researchers in China are addressing these issues by developing a new harvesting robot model focused on 'human–robot collaboration'.
A research team, led by Associate Professor Pei Wang from Southwest University, introduced a gesture-controlled human–robot collaborative harvesting robot that advances the efficiency of fruit picking through an innovative approach. This technology aims to enhance the productivity of small-scale orchards transitioning towards intelligent systems.
The core technology leverages the strengths of both humans and machines: Humans are adept at identifying fruit locations and choosing picking paths, while robotic arms excel in repetitive tasks and force control. The team's motion-sensing interaction system utilizes real-time input from a Leap Motion sensor to capture hand movements, guiding the robotic arm to the correct position. This system combines human visual skills with robotic mechanical precision.
Technical challenges were addressed to ensure precise operation by the robotic arm. Notably, the team resolved issues related to inverse kinematics calculations, which can produce multiple solutions leading to erratic motions. A four-step screening method was developed, incorporating checks for mechanical interference, verifying accuracy, evaluating motion rationality, and optimizing trajectory smoothness.
In contrast to camera-dependent traditional robots, the new robot utilizes motion-sensing technology with the Leap Motion controller, capable of capturing hand movements at a 0.01-millimeter resolution. It maintains stable performance under suboptimal lighting or when foliage occludes. Intelligent filtering algorithms mitigate 'jittery data' from hand tremors or environmental factors, promoting smooth arm movement.
The team effectively mapped the interaction space of Leap Motion to the robotic arm's working area, enabling operators to control the arm within a virtual 'box' with ease. This interface is intuitive, akin to playing a motion-sensing video game, and shows promise in improving efficiency in small-scale orchards, even in complex environments.
Tests demonstrated an average system response time of 74.4 milliseconds, with a 96.7% accuracy rate in recognizing gestures. Post-training, operators reduced the time to pick a single fruit from 8.3 seconds to 6.5 seconds, confirming the system's adaptability to complex terrains and varying orchard conditions.
Source: Food Technology & Manufacturing