以行為切換為基礎之軟式計算方法於未知環境中移動式機器人導航

Behavior-Based Soft Computing Approaches for Mobile Robot Navigation in an Unknown Environment

蔡耀正(Yao-Cheng Tsai)

指導教授 ： 陳政宏

國立虎尾科技大學/電機資訊學院/電機工程研究所/碩士(2015年)

https://doi.org/10.6827/NFU.2015.00143

摘要

本篇論文提出一個新的以行為切換為基礎之軟式計算於未知環境中移動式機器人導航。我們把行為分成三個區塊而分配不同的任務。第一個行為模式是使用多策略蜂群演算法(Multiple Strategy Artificial Bee Colony algorithm, MSABC)為基礎設計出的補償性類神經模糊網路控制器(Compensatory Neuro-Fuzzy Controller, CNFC) 於導航模式上面。第二個行為模式是模糊控制器為基礎的避障沿牆模式用於沿牆避開障礙物。第三個行為模式是使用主成分分析之倒傳遞網路來診斷出特殊的環境。當機器人為導航模式時，CNFC的輸入是距離感測器的值和機器人與目標物的夾角，輸出為機器人兩輪的速度。而適應函數的設計是根據CNFC在導航模式中評估控制器的三個因子。適應函數要評估的因子總共有三個:導航時間、起點到終點的直線距離和機器人往終點前進的總里程。利用這些因子去設計出一個不用訓練資料也可以評估的適應函數。而原始蜂群演算法(ABC)中雖然有好的探索力，但是開發力卻很差的演算法。因此本論文提出一個多策略蜂群演算法來提升原始演算法的效能。我們將多個差分演算法裡的突變策略引入到多策略蜂群演算法並且自適應策略來平衡蜂群演算法的開發和探索問題。當機器人為避障沿牆模式時，模糊邏輯控制器的輸入是距離感測器的值，輸出是機器人兩輪的速度。避障沿牆模式的目的是為了當機器人遇到障礙物或特殊障礙物時，可以用沿牆來避開障礙物。第三個行為模式是使用倒傳遞網路來診斷環境，如果倒傳遞網路診斷出前方為特殊環境，機器人就會在前方建立虛擬牆使機器人不會進入到特殊環境中並切換到第二個行為來延牆避開障礙物。我們利用第二個行為與第三個行為來避障沿牆所有的障礙物。最後本論文將比較多策略演算法、原始蜂群演算法和其它演算法設計補償性模控制器於導航任務上的效能。本論文在實驗部分將新的行為切換為基礎之軟式計算方法應用於真實的移動機器人並在現實未知環境下進行測試並驗證方法的可行性。最後實驗結果證明了使用以行為切換為基礎之軟式計算的方法，機器人可以在遇到特殊環境前建立虛擬牆使機器人不會進入陷阱裡並到達目標物。

關鍵字

導航控制 ； 模糊控制 ； 主成分分析 ； 類神經網路 ； 機器人運動控制 ； 原始蜂群演算法 ； 差分進化演算法

並列摘要

This dissertation proposes novel behavior-based soft computing approaches for mobile robot navigation in an unknown environment. The behavior were divided three types, and each behavior was assigned a unique task. The first behavior was a multiple strategy artificial bee colony (MSABC) algorithm for designing a compensatory neuro-fuzzy controller (CNFC) to complete an actual mobile robot navigation task. For the second behavior, we designed a wall-following fuzzy logic controller for avoiding obstacles, third behavior was principle component analysis back propagation network(PCA-BPN) based fuzzy logic controller, which is diagnostic special environment for mobile robot escape from traps. During the navigation task, the CNFC inputs are the measured distance and angle between the mobile robot and a target, and the outputs of the CNFC are the robot's left and right-wheel speeds. A fitness function was defined to evaluate the performance of the CNFC in the navigation task. The fitness function comprised the following three control factors (CF) : navigation time, the distance between start point and the target, and the distance between the mobile robot and target. The original artificial bee colony algorithm (ABC) simulates the intelligent foraging behavior of honey-bee swarms, which are effective for exploration but ineffective for exploitation. The proposed multiple strategy ABC algorithm(MSABC) adopts the mutation strategies of differential evolution to balance exploration and exploitation. The purpose of the wall-following obstacle-avoiding behavior is ensuring that when the mobile robot encounters an any obstacle, it can move along the wall and avoid the obstacle. The third behavior was designed to assist in evaluating special environment and creates virtual wall. If the mobile robot determine that it is in a special environment, it creates virtual wall and change to wall-following obstacle-avoiding behavior avoid virtual wall. The mobile robot's second and third behaviors are designed to wall-following with avoid obstacles and virtual wall. To demonstrate the performance of the MSABC designed CNFC, the method was compared with other population-based algorithms with respect to the efficiency of the navigation task. To demonstrate the feasibility of the design, experiments carried out on an actual mobile robot (PIONEER 3-DX) are included in this research. In the propose method, we use novel behavior based soft computing approaches classify obstacle and create of the virtual wall make mobile robot more effectively avoid obstacle and reach the target.

並列關鍵字

Navigation control ； fuzzy logic control ； principle component analysis ； back propagation neural network ； robot motion control ； artificial bee colony algorithm ； differential evolution

參考文獻

[35] D. T. Pham, A. Ghanbarzadeh, E. Koc, S. Otri, S. Rahim, and M. Zaidi, The Bees Algorithm, Technical Report, Manufacturing Engineering Centre, Cardiff University, Cardiff CF. 24 3AA, 2005.

[1] M. Mucientes and J. Casillas, “Quick design of fuzzy controllers with good interpretability in mobile robotics,” IEEE Transactions on Fuzzy Systems, vol. 15, no. 4, pp. 636–651, Aug. 2007.

[2] M. Mucientes, J. Alcala-Fdez, R. Alcala, and J. Casillas, “A case study for learning behaviors in mobile robotics by evolutionary fuzzy systems,” Expert Systems with Applications, vol. 37, no. 2, pp. 1471–1493, Mar. 2010.

[3] A. Khalid, M. Amar, A. Habiba, S. Shafique, and R. Noor, “Design and low cost implementation of a fuzzy logic controller for wall following behavior of a mobile robot,” 2010 2nd International Conference on Signal Processing Systems, vol. 2, pp. 740–746, 5-7 Jul. 2010.

[4] J. A. Saleh, F. Karray, M. S. Masmoudi, and M. Masmoudi, “Soft computing techniques in intelligent wall following control for a car-like mobile robot,” 2009 3rd International Conference on Signals, Circuits and Systems, pp. 1–6, 6-8 Nov. 2009.