A NEW APPROACH TO ASSESS HIGH LEVEL PLANNING UNDERLYING COGNITIVE-MOTOR PERFORMANCE DURING COMPLEX ACTION SEQUENCES

Loading...
Thumbnail Image

Files

Publication or External Link

Date

2018

Citation

Abstract

While much work has examined low-level sensorimotor planning, only limited efforts have studied high-level motor planning processes underlying the cognitive-motor performance of complex action sequences. Such sequences can generally be successfully executed in a flexible manner and typically involve few constraints. In particular, no past study has examined the concurrent changes of high-level motor plans along with those of mental workload and confidence during practice of a novel complex action sequence. To address this gap, first a computational approach providing markers capturing performance dynamics of action sequences during practice had to be developed since past relevant works only employed fairly rough metrics. Such an approach should provide concise performance markers (e.g., distances, scalar) while still capturing accurately the changes of structure of high-level motor plans during the acquisition of novel complex action sequences. Thus, by adapting the Levenshtein distance (LD) and its operators to the motor domain, a computational approach was first proposed to assess in detail action sequences during an imitation practice task executed by various performers (humans, a humanoid robot) and with flexible success criteria. The results revealed that this approach i) could support accurately comparing the high-level plans generated between performers; ii) provides performance markers (LD, insertion operator) able to differentiate optimal (using a minimum of actions) from suboptimal (using more than a minimum of actions but still reaching the task goal) sequences; and iii) gives evidenced that the deletion operator is a marker of action sequence failure. This computational approach was then deployed to examine during practice the concurrent changes in high-level motor plans underlying action sequence execution with modulation of mental workload and an individual’s confidence in performing the task. The results revealed that as individuals practiced, performance improved (reduction of LD, insertion/substitution and movement time) while the level of mental workload and confidence decreased and increased, respectively. Also, by late practice the sequences were still suboptimal while being executed faster, possibly suggesting different dynamics between the generation of high-level motor plans and their execution. Overall, this work complements prior efforts to assess complex action sequences executed by humans and humanoid robots in the context of cognitive-motor practice, and it has the potential to inform not only human cognitive-motor mechanisms, but also human-robots interactions.

Notes

Rights