Multi-Objective Trajectory Optimization to Improve Ergonomics in Human Motion

Latent Ergonomics Maps: Real-Time Visualization of Estimated Ergonomics of Human Movements

“…Moreover, we want to leverage the LEM for ergonomics optimization of robot motions to improve human-robot collaboration. The idea is to inform the robot about the ergonomics risk associated with planned collaborative robot trajectories as in [ 44 ], and then to optimize these trajectories using ergonomics optimization as in [ 43 , 45 ].…”

“…The LEM created using the RULA-C score creates, by definition, a latent space that is similar to that of the RULA but continuous. While there is no immediate advantage in using one of the two LEMs for ergonomics visual feedback, the continuous LEM can be used in future work as a reduced model for planning ergonomics movements using gradient-based methods [ 43 ].…”

Human-centered management in polyergatic information systems. Multi-criteria distribution of functions between operators

“…Researchers of the "human factor" problems note a fundamental change [16][17][18][19] in the role of the human operator in automated systems (figure 1). Changing the role of the human operator [18,19].…”

NeuroErgo: A Deep Neural Network Method to Improve Postural Optimization for Ergonomic Human-Robot Collaboration

“…Many researchers attempted to tackle this optimization problem with different ergonomic assessment methods [9]- [12]. Among the ergonomic assessment methods, the standard methods like Rapid Entire Body Assessment (REBA) method [13] and Rapid Upper Limb Assessment (RULA) method [14] have been mostly utilized [12], [15] These studies fitted a second-order polynomial function on the tabular data of REBA and RULA to derive a differentiable model of the methods. Because the tabular methods are discrete functions and computationally make complicated optimization of problems in which utilized them.…”

“…Our novel method, called NeuroErgo, tackles this problem by learning the corresponding tables, REBA or RULA, using a deep feedforward network [16] in a pre-processing phase. Then, similar to existing methods such as [15] and [12], optimize the corresponding objective functions that are using the result of the network as an approximation function. In other words, polynomials in those methods are replaced by the proposed network in NeuroErgo that gives a higher precision approximation from corresponding ergonomic metrics.…”