Embedded AI System of Low-Cost Sensor for Robotic Gripper in the Box Handling Process

Abstract

As the complexity of global supply chain increases, logistics system are required to be flexible and highly automated. However, large distribution centers run on the automated robots integrated with high-cost sensor system while small warehouses that cope with a large number of small-scale volumes rely on the manual work. The one of the critical aspects to be considered during loading task is the weight of baggage packed into boxes is and is measured by expensive built-in sensors. Typically, conveyor belts containing a weight scaler are used for assuring delivery quality in automated robot system supported by remote intelligent controller computer. This system needs too high installation cost to small warehouse. To reduce the cost, this paper proposed a method, sensor-integrated automation robot system is replaced to a low-cost sensory gripper with an attached intelligent module. Addressing this challenge, an innovative solution is proposed through the embedded artificial intelligence of low-cost sensors for a robotic gripper in a box handling process. The experiment of the prototype gripper shows that a quantized neural network is reduced into 11% size to be operated in a microcontroller attached on the robot, obtaining signals from two low-cost force sensing resistors. The resulting intelligent system encourages the use of sensory grippers embedded artificial intelligence in the industry and automation by identifying the weight of the gripping targets without the need for a separate process or expensive equipment.