Faster and Deterministic Subtrajectory Clustering

Given a trajectory $T$ and a distance $\Delta$, we wish to find a set $C$ of curves of complexity at most $\ell$, such that we can cover $T$ with subcurves that each are within Fr\'echet distance $\Delta$ to at least one curve in $C$. We call $C$ an $(\ell,\Delta)$-clustering and aim to find an $(\ell,\Delta)$-clustering of minimum cardinality. This problem was introduced by Akitaya $et$ $al.$ (2021) and shown to be NP-complete. The main focus has therefore been on bicriterial approximation algorithms, allowing for the clustering to be an $(\ell, \Theta(\Delta))$-clustering of roughly optimal size. We present algorithms that construct $(\ell,4\Delta)$-clusterings of $\mathcal{O}(k \log n)$ size, where $k$ is the size of the optimal $(\ell, \Delta)$-clustering. For the discrete Fr\'echet distance, we use $\mathcal{O}(n \ell \log n)$ space and $\mathcal{O}(k n² \log³ n)$ deterministic worst case time. For the continuous Fr\'echet distance, we use $\mathcal{O}(n² \log n)$ space and $\mathcal{O}(k n³ \log³ n)$ time. Our algorithms significantly improve upon the clustering quality (improving the approximation factor in $\Delta$) and size (whenever $\ell \in \Omega(\log n)$). We offer deterministic running times comparable to known expected bounds. Additionally, in the continuous setting, we give a near-linear improvement upon the space usage. When compared only to deterministic results, we offer a near-linear speedup and a near-quadratic improvement in the space usage. When we may restrict ourselves to only considering clusters where all subtrajectories are vertex-to-vertex subcurves, we obtain even better results under the continuous Fr\'echet distance. Our algorithm becomes near quadratic and uses space that is near linear in $n \ell$.