Arboreal animals often leap through complex canopies to travel and avoid predators. Their success at making split-second potentially life-threatening decisions of biomechanical capability depends on skillful use of acrobatic maneuvers and learning from past efforts. Here, we found that free-ranging fox squirrels (Sciurus niger) leaping across novel, simulated branches decided where to launch by balancing a trade-off between gap distance and branch-bending compliance. Squirrels quickly learned to modify impulse generation upon repeated leaps from novel, compliant beams. A repertoire of agile landing maneuvers enabled targeted leaping without falling. Unanticipated adaptive landing and leaping “parkour” behavior revealed an innovative solution for, particularly challenging leaps. Squirrels deciding and learning how to launch and land demonstrate the synergistic roles of biomechanics and cognition in robust gap crossing strategies.

Experiment 1 - Launch-point decision. All trials were recorded with two Phantom V10 high speed cameras (Vision Research Inc) at 200 frames per second using 50mm lenses and Phantom Camera Control 1.3 software. One camera was placed such that it had a lateral view of squirrels to record their movements in the sagittal plane, and bisected the distance jumped by the squirrel during tests. It was also level with the jumping platform and framed such that only the overhanging beam and the landing perch were in frame. The second camera was placed to record the squirrels’ movements in the coronal plane. Twelve squirrels were recorded in 96 leaping trials. The number of trials per individual was 23, 27, 3, 19, 4, 4, 2, 1, 2, 4, 2, and 5.

The data sheet contains metadata and analysis for each trial.
DataSheetExperiment1.csv
DataSheetExperiment1.xlsx

Experiment1/AnalysisScripts
The primary analysis script is launch_point_decisions.m. This MATLAB script uses the data in launch_points.xlsx to analyze the launch point decisions from each trial. The script force_displacement.m uses the data from testing_branch_compliance.xlsx to determine the local compliance along each simulated branch. The other two scripts, linspecer.m and plot2axes, are called by the launch_point_decisions.m script and used for plotting.

Experiment 2 - Learning to leap. A single Phantom V10 camera recorded sagittal plane movements of the squirrel at 200 frames per second. In total, we recorded 47 trials from 5 squirrels. Four squirrels completed 10 leaps, 5 from the highly flexibly beam and 5 from the rigid control beam. A fifth squirrel completed 8 trials including all 5 leaps from the flexibly beam, but left the area after 3 leaping trials from the rigid control beam. Times between leaping trials from the flexible beam varied from 33 to 197 seconds with an average of 109 seconds.

The data sheet contains metadata and analysis for each trial.
DataSheetExperiment2.csv
DataSheetExperiment2.xlsx

Experiment2/AnalysisScripts
The primary analysis script is squirrel_datasheet_generator.m. This MATLAB script uses the tracked aerial phases to calculate the landing errors and body curvatures. The script squirrel_tracker is used to manually track body positions on the squirrels from the video data and output a tracked data file. The script squirrel_COM is called by squirrel_datasheet_generator.m, and uses the tracked data to estimate the center of mass position and landing error.

Experiment2/AnalysisScripts/tracked_aerial_phase.zip
The manually tracked kinematics are stored in separate .mat files (MATLAB) for each leaping trial.

Experiment 3 - Parkour leaping. One camera was placed to record squirrel movements in the sagittal plane Another camera was mounted above the wall apparatus so that it had a downward view of the squirrels. Lens distortions were corrected prior to kinematic analysis using custom MATLAB code and the native functions estimateCameraParameters, and undistortImage. Squirrels were presented with a single landing perch at one of the eight locations (Fig. 3B) in a randomized order. In total, 324 parkour leaping trials were recorded from ten individual squirrels. The number of leaping trials per individual was 31, 47, 31, 8, 9, 32, 51, 54, 41, and 20. The center of mass was tracked.

Experiment 3/AnalysisScripts
The primary analysis script is calculate_velocity_before_and_after_wall_jump.m. This MATLAB script uses the tracked video data to calculate velocities during the aerial phases before and after the wall contact phase of each leaping trial. The script struct2vec.m is used to reformat the data structure from the tracked video data. This script is called by calculate_velocity_before_and_after_wall_jump.m. The script calculate_COM.m is used to estimate the center of mass from the tracked video data. This script is called by calculate_velocity_before_and_after_wall_jump.m.

Experiment3/AnalysisScripts/tracked_video_data.zip
The manually tracked kinematics are stored in separate .mat files (MATLAB) for each leaping trial.