The invention relates to a method for testing palletizing robot control software based on Agent, and the method provided by the invention is an improved method for carrying out black box testing and white box testing on the palletizing robot control software, and belongs to the field of embedded software testing. The method provided by the invention mainly comprises the following four steps: 1, building an Agent-based BDI (belief-desire-intention) model TC_Agent, and generating a black box testing case required for testing; 2, judging the implementation situation of the systemic function by using a black box testing Agent; 3, building a genetic algorithm-based coverage testing case generation model GA_Agent, and realizing the automatic generation of coverage testing cases through carrying out a series of genetic operations such as code design, fitness function design, selection, crossover, mutation, and the like; and 4, building a white box testing Agent model, completing the instrumentation operation of programs through carrying out preprocessing and program instrumentation on Agent, compiling and running the programs subjected to instrumentation, then sending the programs to a target machine, inputting the testing cases generated by the GA_Agent, and analyzing the Agent by using a coverage rate so as to obtain the coverage rates of program statements and branches of a palletizing robot. The method disclosed by the invention is simple and quick, is strong in commonality, can be applied to different programs, and can be used for greatly improving the efficiency of the software automation test.