<p>Walking tractors (WTs) are gaining popularity globally due to their maneuverability and suitability for small-scale farms. However, traditional WTs lack dedicated steering mechanisms and rely on a single clutch for power transmission. This study proposes a novel cone clutch design to improve maneuverability by facilitating turns during operation. Due to space and weight constraints on the WT's turning axle, the clutch design prioritizes minimal weight while maintaining necessary dimensions. Machine learning techniques, including Response Surface Methodology (RSM) with Central Composite Design (CCD) and Artificial Neural Network (ANN), were employed to optimize cone clutch component weights (outer cone, inner cone, and friction lining). The central composite design (CCD) has been employed to determine the effect of various parameters namely clutch dimensions on weight and stress. ANN effectively predicted weight using RSM data (correlation coefficient of 0.99, 80% data split). The final design was optimized through a Genetic Algorithm (GA) coupled with Finite Element Analysis (FEA) for static, dynamic, and thermal analyses. This comprehensive approach ensured the clutch's stability under various load conditions. This research demonstrates the effectiveness of combining RSM, MLR, ANN, GA, and FEA for designing a lightweight, dimensionally-constrained cone clutch for enhanced maneuverability in WTs. This optimized design offers a promising solution for improving steering capabilities in walking tractors.</p> Graphical Abstract <p></p>