A CNN pallet detection model for an autonomous forklift.

Autonomous mobile robots are used in industrial applications, mainly for warehouse logistics and transport. This equipment can transport light and heavy loads, and operate practically 24 hours a day, seven days a week, with short recharge stops. A new challenge for these devices is cargo collection. Some industries use pallets to store their products, so the robot must collect them autonomously. For this, the robot must be able to identify the pallet and fill it with the appropriate cargo. This work approaches pallet identification through point cloud processing provided by the LiDAR sensor, which is present in practically every autonomous mobile robot. This work develops A Convolutional Neural Network pallet detection model for an autonomous forklift. The model is trained and evaluated with CoppeliaSim and reached a good performance for simulated data.