This study focuses on learning of control commands identification and load identification for active shape control of a tensegrity structure in situations of unknown loading event. Control commands are defined as sequences of contractions and elongations of active struts. Case-based reasoning strategies support learning. Simple retrieval and adaptation functions are proposed. They are derived from experimental results of load identification studies. The proposed algorithm leads to two types of learning: reduction of command identification time and increase of command quality over time. In the event of no retrieved case, load identification is performed. This methodology is based on measuring the response of the structure to current load and inferring the cause. It provides information in order to identify control commands through multi-objective search. Results are validated through experimental testing.