Researchers from the Dept. of Automatics, Biocybernetics, and Robotics at Jožef Stefan Institute have developed a new method for autonomous learning of disassembly tasks with robots. While this type of learning is often very time-consuming, the proposed approach enables the robot to quickly discover an optimal solution. The basis for rapid searching is the consideration of physical constraints that arise while dismantling objects, allowing the robot to significantly reduce the search space. The authors of the proposed method, Mihael Simonič, Aleš Ude, and Bojan Nemec, published it in the paper titled Hierarchical learning of robotic contact policies, which has recently been published in a prestigious journal Robotics and Computer-Integrated Manufacturing. This research was conducted in the frame of Horizon 2020 project (ReconCycle), which is coordinated by prof. Ude. The developed method will contribute to the faster development of applications for robotic recycling of electronic devices, which is only possible with highly adaptive robots. The high degree of adaptivity is necessary due to the variability of electronic devices.