本研究以可回溯式門檻接受法(Backtracking Adaptive Threshold Accepting, BATA)為基本架構,並與一般化插入/解繫法(GENIUS)和其他方法相結合,求解VRP問題。測試例題為Christofide et al. (1979)之14題標竿題目。結果顯示,依傳統BATA設定門檻回溯比值b<1時,14題平均誤差可為1.2%。而若突破此限制,設定b>1時,則最低為0.87%,表示其求解品質可能更佳。本研究各題最佳結果之平均誤差亦僅有0.26%。 Backtracking Adaptive Threshold Accepting (BATA) is similar to Threshold Accepting (TA) but the values of threshold are lowered or raised, depending on if an acceptable solution can be found in a fixed number of iterations. This research used a BATA struc