Work

Assessment of warning modalities for remotely supervised autonomous agricultural machines

Year: 2025
Type: article
Abstract: As autonomous systems gain traction in agriculture, effective human-computer interaction (HCI) frameworks are essential for safe and efficient remote supervision. While autonomous agricultural machines (AAMs) improve productivity, human operators remain vital for monitoring, interpreting warnings, and intervening during system failures. The design of warning systems within human-machine interfaces (HMIs) directly affects operator performance and situation awareness (SA), including perception (Level 1), comprehension (Level 2), and projection (Level 3). This research investigates how visual, auditory, and tactile warning modalities, used alone or in combination, influence operator awareness and responsiveness during remote AAM supervision. Through simulated field scenarios, participants engaged with unimodal and bimodal cues, with objective measures (reaction time, comprehension, and projection accuracy, CSUQ usability scores) and subjective ratings collected and analyzed. Bimodal cues enhanced perception. Visual-tactile warnings led to shorter response times than visual-auditory cues in quiet (close-to-field) environments, though no significant difference was found under noisy (in-field) conditions. Both modalities received high usability ratings, with farming participants preferring visual-tactile cues. Among non-farming participants, response times were significantly faster with visual-auditory warnings, t(24) = 3.16, p = .004 (saccadic), and t(24) = 3.56, p = .002 (manual). Projection accuracy was lower for farming participants using visual-auditory cues (86.7%), though not statistically significant, χ²(3, N = 540) = 5.37, p = .147. Among unimodal cues, visual warnings were most effective across comprehension, urgency perception, and response time, earning the highest user preference (84%). Auditory cues showed moderate performance; tactile cues were least effective. Correlation analysis showed faster comprehension and predicted higher usability, with task accuracy aligning with CSUQ ratings. Together the findings support the development of adaptive, user-centered HMIs that align warning modality with cognitive and environmental demands. The results of this research contribute to safer, more effective human-robot collaboration in agricultural automation.
Language: English
Open Access status: green
OpenAlex ID: W7137452929
FWCI:
Cites: 168
Cited by:
Related to: 10