Backtracking is when a search or related algorithm returns to a previous state. This may be because the algorithm has got stuck, for example when a tree search encounters a node that is infeasible, that is does not satisfy all of the constraints for an acceptable solution. Alternatively, when the algorithm has found some form of a solution, it may want to look for more. Typically an algorithm backtracks to a state where there were multile choices of what to do next and then takes another path from there.
Used in Chap. 2: pages 16, 23; Chap. 4: pages 44, 48, 50, 54; Chap. 15: page 224