The traveling salesman problem (TSP) has been a popular subject for genetic algorithm research. The heavy constraints of the problem make the results of a standard crossover operation invalid. This paper investigates a technique of chromosome repair called Directed Edge Repair (DER) that compensates for invalid edges from crossover or mutation. The algorithm is combined with the technique of brood selection from genetic programming and with the 3-opt local search technique. Tests on several problems from TSPLIB are reported. The algorithm was able to find optimal solutions to problems up to 1577 cities. On many of the problems, brood selection improved the accuracy of solutions without adding additional computation time.