An Analytic Model for Negative Obstacle Detection with Lidar and Numerical Validation Using Physics-Based Simulation.

Negative obstacles have long been a challenging aspect of autonomous navigation for ground vehicles. However, as terrestrial lidar sensors have become lighter and less costly, they have increasingly been deployed on small, low-flying UAV, affording an opportunity to use these sensors to aid in autonomous navigation. In this work, we develop an analytical model for predicting the ability of UAV or UGV mounted lidar sensors to detect negative obstacles. This analytical model improves upon past work in this area because it takes the sensor rotation rate and vehicle speed into account, as well as being valid for both large and small view angles. This analytical model is used to predict the influence of velocity on detection range for a negative obstacle and determine a limiting speed when accounting for vehicle stopping distance. Finally, the analytical model is validated with a physics-based simulator in realistic terrain. The results indicate that the analytical model is valid for altitudes above 10 m and show that there are drastic improvements in negative obstacle detection when using a UAV-mounted lidar. It is shown that negative obstacle detection ranges for various UAV-mounted lidar are 60-110 m, depending on the speed of the UAV and the type of lidar used. In contrast, detection ranges for UGV mounted lidar are found to be less than 10 m.

Semantic Segmentation with Transfer Learning for Off-Road Autonomous Driving.

Since the state-of-the-art deep learning algorithms demand a large training dataset, which is often unavailable in some domains, the transfer of knowledge from one domain to another has been a trending technique in the computer vision field. However, this method may not be a straight-forward task considering several issues such as original network size or large differences between the source and target domain. In this paper, we perform transfer learning for semantic segmentation of off-road driving environments using a pre-trained segmentation network called DeconvNet. We explore and verify two important aspects regarding transfer learning. First, since the original network size was very large and did not perform well for our application, we proposed a smaller network, which we call the light-weight network. This light-weight network is half the size to the original DeconvNet architecture. We transferred the knowledge from the pre-trained DeconvNet to our light-weight network and fine-tuned it. Second, we used synthetic datasets as the intermediate domain before training with the real-world off-road driving data. Fine-tuning the model trained with the synthetic dataset that simulates the off-road driving environment provides more accurate results for the segmentation of real-world off-road driving environments than transfer learning without using a synthetic dataset does, as long as the synthetic dataset is generated considering real-world variations. We also explore the issue whereby the use of a too simple and/or too random synthetic dataset results in negative transfer. We consider the Freiburg Forest dataset as a real-world off-road driving dataset.

Closing the Wearable Gap-Part V: Development of a Pressure-Sensitive Sock Utilizing Soft Sensors.

The purpose of this study was to evaluate the use of compressible soft robotic sensors (C-SRS) in determining plantar pressure to infer vertical and shear forces in wearable technology: A ground reaction pressure sock (GRPS). To assess pressure relationships between C-SRS, pressure cells on a BodiTrakTM Vector Plate, and KistlerTM Force Plates, thirteen volunteers performed three repetitions of three different movements: squats, shifting center-of-pressure right to left foot, and shifting toes to heels with C-SRS in both anterior-posterior (A/P) and medial-lateral (M/L) sensor orientations. Pearson correlation coefficient of C-SRS to BodiTrakTM Vector Plate resulted in an average R-value greater than 0.70 in 618/780 (79%) of sensor to cell comparisons. An average R-value greater than 0.90 was seen in C-SRS comparison to KistlerTM Force Plates during shifting right to left. An autoregressive integrated moving average (ARIMA) was conducted to identify and estimate future C-SRS data. No significant differences were seen in sensor orientation. Sensors in the A/P orientation reported a mean R2 value of 0.952 and 0.945 in the M/L sensor orientation, reducing the effectiveness to infer shear forces. Given the high R values, the use of C-SRSs to infer normal pressures appears to make the development of the GRPS feasible.

Impact of military type footwear and load carrying workload on postural stability.

Postural stability has been shown to be impacted by footwear and task performed. This study analysed the impact of two military footwear, standard boot (STB) and minimalist boot (MTB) on postural stability, before (PRE) and after (POST) a load carriage task. Sixteen participants were tested for postural stability using sensory organisation and motor control tests on Neurocom Equitest™. Postural sway, equilibrium scores and postural latencies were analysed using a two-factor repeated measures ANOVA: boot type (STB-MTB) × time (PRE-POST) load carriage task. Significantly greater postural sway variables, lower balance scores and slower postural latencies were seen in STB and POST load carriage conditions (p < .05). The results suggest that MTB exhibited greater balance compared to STB in balance conditions that rely on somatosensory feedback and that balance is lowered after a load carriage task. Decrements in postural stability could be attributed to boot design characteristics and muscular exertion due to the load carriage task. Practitioner Summary: Maintaining optimal postural stability is crucial in military. Impact of military footwear types and load carriage task on postural stability are addressed. Findings provide footwear design and physical exertion implications on postural stability leading to potential interventions that reduce postural stability decrements; thereby, reducing potential falls and fall related injuries.

Impact of Virtual Reality-Generated Construction Environments at Different Heights on Postural Stability and Fall Risk.

BACKGROUND: Falls due to postural instability are common in construction environments especially from a height. The purpose of the study was to investigate the impact of virtual reality (VR)-generated environments at different virtual heights on postural stability. METHODS: Nineteen adults were analyzed for postural stability, tested in real (No VR) environment and in three VR environments, randomly assigned, at virtual heights of 0 ft. (VR0), 40 ft. (VR40), and 120 ft. (VR120). Postural stability was quantified using center of pressure postural sway variables and analyzed using a repeated measures analysis of variance (ANOVA). Participants also completed a simulation sickness questionnaire (SSQ) before and after VR exposure and a presence questionnaire (PQ) after VR exposure. FINDINGS: Significant postural instability (p < .05) was identified between VR and No VR, in which increased postural instability was evident in all VR conditions compared with No VR. Scores from SSQ were within a pre-post score difference of five and the PQ score was (104.21 ± 14.03). CONCLUSION/APPLICATION TO PRACTICE: Findings suggest that VR environments, regardless of virtual height, induced increased postural instability, which can be attributed to visual sensory conflicts to the postural control system created by VR exposure. Participants' subjective responses on SSQ and PQ confirmed the feasibility of using VR to represent realistic immersions in virtual heights. However, objectively, VR could potentially lead to postural instability, stressing caution. VR can be a potential tool for providing virtual high-altitude environment exposure for fall prevention training, however, more research is needed on postural adaptation with acute and chronic exposure to VR.

Virtual-Reality-Induced Visual Perturbations Impact Postural Control System Behavior.

BACKGROUND: Virtual reality (VR) is becoming a widespread tool in rehabilitation, especially for postural stability. However, the impact of using VR in a "moving wall paradigm" (visual perturbation), specifically without and with anticipation of the perturbation, is unknown. METHODS: Nineteen healthy subjects performed three trials of static balance testing on a force plate under three different conditions: baseline (no perturbation), unexpected VR perturbation, and expected VR perturbation. The statistical analysis consisted of a 1 × 3 repeated-measures ANOVA to test for differences in the center of pressure (COP) displacement, 95% ellipsoid area, and COP sway velocity. RESULTS: The expected perturbation rendered significantly lower (p < 0.05) COP displacements and 95% ellipsoid area compared to the unexpected condition. A significantly higher (p < 0.05) sway velocity was also observed in the expected condition compared to the unexpected condition. CONCLUSIONS: Postural stability was lowered during unexpected visual perturbations compared to both during baseline and during expected visual perturbations, suggesting that conflicting visual feedback induced postural instability due to compensatory postural responses. However, during expected visual perturbations, significantly lowered postural sway displacement and area were achieved by increasing the sway velocity, suggesting the occurrence of postural behavior due to anticipatory postural responses. Finally, the study also concluded that VR could be used to induce different postural responses by providing visual perturbations to the postural control system, which can subsequently be used as an effective and low-cost tool for postural stability training and rehabilitation.

Impact of occupational footwear and workload on postural stability in work safety.

BACKGROUND: The impact of occupational footwear and workload on postural stability has been studied previously to prevent fall-related workplace injuries. OBJECTIVE: The purpose of this study was to assess the impact of two types of occupational footwear [steel-toed (SB) and tactical (TB) work boots] on human balance, when exposed to physical workload. METHODS: Postural stability was evaluated in eighteen male participants in the following conditions: eyes open (EO), eyes closed (EC), eyes open unstable surface (EOU) and eyes closed unstable surface (ECU). Postural sway parameters were analyzed using a 2×3 repeated measures analysis of variance design [prior to (PRE) and twice post-workload (POST1 & POST2) separated by 10 minutes of rest]. RESULTS: Findings revealed that the use of SB resulted in greater postural stability, which could be attributed to the design characteristics of these footwear and that postural stability was negatively impacted immediately after the workload which could be attributed to the physical exertions during the workload. However, significant differences were limited to ECU with no visual and altered somatosensory feedback. CONCLUSION: Design features on occupational footwear can aid postural stability while physical exertional tasks can be detrimental. Findings can offer design and work-rest scheduling suggestions to improve work safety.

Efficacy of virtual reality in pedestrian safety research.

Advances in virtual reality technology present new opportunities for human factors research in areas that are dangerous, difficult, or expensive to study in the real world. The authors developed a new pedestrian simulator using the HTC Vive head mounted display and Unity software. Pedestrian head position and orientation were tracked as participants attempted to safely cross a virtual signalized intersection (5.5 m). In 10% of 60 trials, a vehicle violated the traffic signal and in 10.84% of these trials, a collision between the vehicle and the pedestrian was observed. Approximately 11% of the participants experienced simulator sickness and withdrew from the study. Objective measures, including the average walking speed, indicate that participant behavior in VR matches published real world norms. Subjective responses indicate that the virtual environment was realistic and engaging. Overall, the study results confirm the effectiveness of the new virtual reality technology for research on full motion tasks.

Evaluating pedestrian behavior at crosswalks: Validation of a pedestrian behavior questionnaire for the U.S. population.

The aim of this study was to develop and validate a self-reporting Pedestrian Behavior Questionnaire (PBQ) for the U.S. population to measure frequency of risky behaviors among pedestrians. The PBQ includes 50 survey items that allow respondents to rate the frequency with which they engage in different types of road-using behaviors as pedestrians. The validation study was conducted on 425 participants (228 males and 197 females) between the ages of 18 and 71. Confirmatory factor analysis differentiated pedestrian behaviors into five factor categories: violations, errors, lapses, aggressive behaviors, and positive behaviors. A short version of the PBQ with 20 items was also created by selecting four items with high factor loadings from each of the five factor categories. Regression analyses investigated associations with scenario-based survey behavioral responses to validate the five-factor PBQ subscale scores and composite score. For both long and short versions, each of these five individual factor scales were found to be reliable (0.7