Background: Accessing targets through complex trajectories with multiple curves in series, remains challenging today in many application fields, like surgery, the Fukushima disaster, maintenance of complex tube networks, mining and space operations. Current steerable devices require control of an extensive amount of segments, while moving as a whole through the environment. In contrast, growth-from-the-tip devices move only at the tip by creating a solidified support structure, allowing for active curvature creation at any point of the trajectory with limited effect on the environment. However, reversible and accurate growth remain challenging for these type of devices.Method: Competitive design requirements are setup and quantified. Subsequently, a broad concept generation is performed, based on the ACRREX (i.e. Abstracting, Categorizing, Reflecting, Reformulating and Extending ) method. After selection of one working principle, a proof-of-principle is designed, prototyped and evaluated. Based on the outcome, a second prototype is created and evaluated again. Evaluation of both prototypes comprise accuracy measurements, based on the ratio of path deviation from a kinematic model and insertion length. Moreover, load bearing capacity of the support structure, forces acting on the environment and overall performance are assessed.Results: The selected concept is based on everting chains, driven by torque at the tip and consisting of two chains with lockable sliding hinged joints. Both prototypes successfully show reversible and steerable growth-from-the-tip, by the fact that counter rotation of the tip gears resulted in a translation of the tip, while differential rotation of the gears resulted in curvature creation. Prototype I (PI) had an accuracy of 4.5% for pure translations and 6.2% to 8.1% for the formation of two sequential (90 degrees) curvatures. Prototype II (PII), which was not fully operational, had a deviation per insertion length of 2.9% for a single curvature and to 10.7% to 19.1% for a combination of translations and curvatures in series. Even though the shape-locks of Prototype I had 3 degrees play per joint and a limiting amount of locking positions, the average orthogonal load bearing capacity was high (49N) with an accompanied deformation of maximally 8 degrees. In contrast, the friction-locked drum brakes of Prototype II rotated up to 70 degrees, by both deformation (22%) and slippage (78%) at a load of 20N. A rotation of 25 degrees was reached at an average load of 7.3N. Normal forces acting on the environment during movement were only measured for Prototype I, resulting in direct distortions (78%) of <4N and indirect distortions (22%) of <0.36N. Moreover, both prototypes had comparable elongation rates (PI: 92.9mm/min, PII: 156mm/min), extension rates (PI: 2.6mm/mm, PII: 0.9mm/mm) and widths (PI: 268mm, PII: 270mm). Lastly, a minimal inner curvature of zero and a minimal outer curvature equal to the device’s width were realized, due to the presence of sliding hinged joints.Conclusion and discussion: Both prototypes performed reversible growth-from-the-tip based on the everting chains principle. Accordingly, the generated kinematic model that links gear rotations to tip rotations and translations seems to be an accurate simplification of the experimental data.In addition, the normal forces acting on the environment were much lower than the load bearing capacity of Prototype I, confirming sufficient support and limited deflection and deformation by the built structure. The everting chain principle has many advantages. First, multiple curves in series can be formed in a reversible manner. Moreover, high load bearing capacity of locked chains results in high accuracy of the system’s movement. Furthermore, the system has limited interaction with the environment and no additional system is required to create curvature. Lastly, sliding hinged joints allow for movement along sharp edges. The next generation of everting chain robots should incorporate improved power and load tuning, size reduction and chains that allow for fully selfsupporting movements in 3D space.