The conventional use of inverse dynamics applied to gait analysis involves the estimation of joint forces and moments based on kinematic data, anthropometric parameters and force plate data. The procedure uses the measured ground reactions and, beginning with those segments in contact with the ground, calculates joint forces and moments at each successive segment. However, this procedure cannot always be applied. There are laboratories that do not have force platforms, or situations in which measurements must be made outside the laboratory. Force plate data only gives the ground forces and moments resultants. When using multi-segment foot models, additional assumptions must be made to distribute the resultants among the different segments in contact with ground. In these situations, a possible solution is the use of inverse dynamics based only on kinematics. This procedure usually starts with the non-contact segments (head, hands, swinging foot in single support phase) and ends with the foot or feet in contact. In the double-support phase of gait, the ground reaction forces include 12 unknowns, which makes the inverse dynamics problem indeterminate. To solve the indeterminacy, the Smooth Transition Assumption (STA) was used in this work. This algorithm is based on the assumption that the reaction forces and moments at the trailing foot decay according to a certain law that combines exponential and linear functions, reducing the number of unknowns to six. In this work, ground reaction force and moments calculated using different body pose reconstruction methods were compared to data provided by force plates. Results depend greatly on the body pose reconstruction method used. Results show that the imposition of kinematic constrains can lead to worse results in the kinetic results if the multibody model is too simplistic. To get good comparison results between inverse dynamics methods based only on kinematics and classical procedures using force plates data, dynamic residuals should be as small as possible. In the single support phase, differences will be identically equal to the dynamic residuals, but higher in the double support phase because of the errors introduced by the transition functions.

逆向动力学应用于步态分析的常规用途包括根据运动学数据，人体测量学参数和测力板数据估算关节力和力矩。该程序使用测得的地面反作用力，并从那些与地面接触的部分开始，计算每个连续部分的联合力和力矩。但是，此过程并不总是适用。有些实验室没有测力平台，或者必须在实验室外进行测量的情况。测力板数据仅给出地面力和力矩合力。当使用多段脚模型时，必须做出其他假设以将结果分配到与地面接触的不同段之间。在这些情况下，一个可能的解决方案是仅基于运动学使用逆动力学。此过程通常从非接触部分（头部，手，单支撑阶段的摆动脚）开始，并在脚接触时结束。在步态的双支撑阶段，地面反作用力包括12个未知数，这使得逆动力学问题不确定。为了解决不确定性，在这项工作中使用了“平滑过渡假设”（STA）。该算法基于以下假设：根据结合了指数函数和线性函数的定律，后脚的反作用力和力矩会衰减，从而将未知数减少到六个。在这项工作中，将使用不同体位重构方法计算的地面反作用力和力矩与测力板提供的数据进行了比较。结果很大程度上取决于所使用的身体姿势重建方法。结果表明，如果多体模型过于简单，则施加运动约束会导致动力学结果更差。为了在仅基于运动学的逆动力学方法和使用力板数据的经典方法之间获得良好的比较结果，动态残差应尽可能小。在单支撑阶段，差异将等同于动态残差，但在双支撑阶段，差异会更大，因为过渡函数会引入误差。为了在仅基于运动学的逆动力学方法和使用力板数据的经典方法之间获得良好的比较结果，动态残差应尽可能小。在单支撑阶段，差异将等同于动态残差，但在双支撑阶段，差异会更大，因为过渡函数会引入误差。为了在仅基于运动学的逆动力学方法和使用力板数据的经典方法之间获得良好的比较结果，动态残差应尽可能小。在单支撑阶段，差异将等同于动态残差，但在双支撑阶段，差异会更大，因为过渡函数会引入误差。