This paper presents a reactive planning system that allows a Cassie-series bipedal robot to avoid multiple non-overlapping obstacles via a single, continuously differentiable control barrier function (CBF). The overall system detects an individual obstacle via a height map derived from a LiDAR point cloud and computes an elliptical outer approximation, which is then turned into a CBF. The QP-CLF-CBF formalism developed by Ames et al. is applied to ensure that safe trajectories are generated. Liveness is ensured by an analysis of induced equilibrium points that are distinct from the goal state. Safe planning in environments with multiple obstacles is demonstrated both in simulation and experimentally on the Cassie biped.