Robots can adapt to damage like animals by using an intelligent trial-and-error algorithm. This method allows robots to quickly find compensatory behaviors without pre-programmed plans or self-diagnosis. The algorithm creates a detailed map of high-performing behaviors before deployment, which guides the robot in learning new behaviors after damage. Experiments show that this approach works for legged robots with damaged, broken, or missing legs and for robotic arms with broken joints. The technique enables more robust and effective autonomous robots, similar to how animals adapt to injuries. Robots are valuable in industries like manufacturing, search and rescue, disaster response, healthcare, and transportation. However, their fragility limits their use in complex environments. Current damage recovery methods rely on self-diagnosis and pre-designed contingency plans, which are costly and often fail. In contrast, injured animals learn through trial and error, adapting creatively to injuries. However, state-of-the-art learning algorithms are impractical due to the "curse of dimensionality." The proposed method uses a behavior-performance map, created with a novel algorithm and simulation, to guide the robot in finding effective behaviors. The map predicts the performance of thousands of behaviors, and the robot uses this information to test and update its predictions. This process allows the robot to adapt quickly and effectively, even without detailed knowledge of the damage's cause. The algorithm uses a Gaussian process model and Bayesian optimization to search for the maximum performance. It balances exploration and exploitation to find the most effective behaviors. The method was tested on a hexapod robot and a robotic arm, showing rapid adaptation and efficient recovery from damage. The results demonstrate that the intelligent trial-and-error approach enables robots to adapt quickly and effectively, similar to animals. This approach has potential applications in various robotics tasks, including adapting to new environments.Robots can adapt to damage like animals by using an intelligent trial-and-error algorithm. This method allows robots to quickly find compensatory behaviors without pre-programmed plans or self-diagnosis. The algorithm creates a detailed map of high-performing behaviors before deployment, which guides the robot in learning new behaviors after damage. Experiments show that this approach works for legged robots with damaged, broken, or missing legs and for robotic arms with broken joints. The technique enables more robust and effective autonomous robots, similar to how animals adapt to injuries. Robots are valuable in industries like manufacturing, search and rescue, disaster response, healthcare, and transportation. However, their fragility limits their use in complex environments. Current damage recovery methods rely on self-diagnosis and pre-designed contingency plans, which are costly and often fail. In contrast, injured animals learn through trial and error, adapting creatively to injuries. However, state-of-the-art learning algorithms are impractical due to the "curse of dimensionality." The proposed method uses a behavior-performance map, created with a novel algorithm and simulation, to guide the robot in finding effective behaviors. The map predicts the performance of thousands of behaviors, and the robot uses this information to test and update its predictions. This process allows the robot to adapt quickly and effectively, even without detailed knowledge of the damage's cause. The algorithm uses a Gaussian process model and Bayesian optimization to search for the maximum performance. It balances exploration and exploitation to find the most effective behaviors. The method was tested on a hexapod robot and a robotic arm, showing rapid adaptation and efficient recovery from damage. The results demonstrate that the intelligent trial-and-error approach enables robots to adapt quickly and effectively, similar to animals. This approach has potential applications in various robotics tasks, including adapting to new environments.